diff --git a/outputs/2026-05-31/16-33-10/.hydra/config.yaml b/outputs/2026-05-31/16-33-10/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..37f20840bdb620aecfcd1d64574488562bcf3127 --- /dev/null +++ b/outputs/2026-05-31/16-33-10/.hydra/config.yaml @@ -0,0 +1,235 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 8 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 700 + customized_validation: true + memory_condition_length: 0 + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + memory_condition_length: 0 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + debug_force_all_streams: false + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + token_patch_size: 2 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 3 + dynamic: + enabled: true + exclude_latest_local_frames: 4 + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + high_quality_fov_threshold: 0.7 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 3 + stage_policy: + noise_bucket_logging: true + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + dit_hidden_size: 1024 + anchor_gate: 1.0 + dynamic_gate: 1.0 + revisit_gate: 1.0 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: DeMemWM + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_full_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_full_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-05-31/16-33-10/.hydra/hydra.yaml b/outputs/2026-05-31/16-33-10/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fd70e7cc2cf00438022c75d6c5c11debb1b0b5ff --- /dev/null +++ b/outputs/2026-05-31/16-33-10/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_full_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=700 + - +dataset.customized_validation=true + - +dataset.memory_condition_length=0 + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_force_all_streams=false + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=3 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=3 + - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=8 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=3,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_force_all_streams=false,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.exclude_latest_local_frames=4,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=3,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.stage_policy.noise_bucket_logging=true,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.memory_condition_length=0,+dataset.n_frames_valid=700,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_full_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_full_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=8,experiment.validation.val_every_n_step=2000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-05-31/16-33-10 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-05-31/16-33-10/.hydra/overrides.yaml b/outputs/2026-05-31/16-33-10/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..47fd8dfd4320154e59dcf5d378b89f43446db9b2 --- /dev/null +++ b/outputs/2026-05-31/16-33-10/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_full_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=700 +- +dataset.customized_validation=true +- +dataset.memory_condition_length=0 +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_force_all_streams=false +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=3 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=3 +- ++algorithm.dememwm.stage_policy.noise_bucket_logging=true +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=8 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-05-31/16-33-10/main.log b/outputs/2026-05-31/16-33-10/main.log new file mode 100644 index 0000000000000000000000000000000000000000..4988357ba5cf33d78d57e07abfaf6c1774e7d009 --- /dev/null +++ b/outputs/2026-05-31/16-33-10/main.log @@ -0,0 +1,38 @@ +[2026-05-31 16:34:05,141][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-05-31 16:34:13,901][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-05-31 16:34:18,601][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-05-31 16:34:18,897][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-05-31 16:34:19,141][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-05-31 16:34:19,288][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-05-31 16:34:34,883][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-05-31 16:34:34,883][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-05-31 16:34:34,883][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-05-31 16:34:34,883][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-05-31 16:34:34,883][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-05-31 16:34:34,883][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-05-31 16:34:39,689][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-05-31 16:34:39,689][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-05-31 16:34:39,689][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-05-31 16:34:39,689][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-05-31 16:34:39,689][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-05-31 16:34:39,689][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-03 16:33:00,182][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-03 16:33:00,183][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-03 16:33:00,184][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-03 16:33:00,182][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-03 16:33:00,184][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-03 16:33:00,184][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-03 16:33:00,183][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-03 16:33:00,186][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-03 16:33:00,190][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-03 16:33:01,747][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-03 16:33:01,748][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-03 16:33:01,748][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-03 16:33:01,749][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-05-31/16-33-11/.hydra/config.yaml b/outputs/2026-05-31/16-33-11/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..37f20840bdb620aecfcd1d64574488562bcf3127 --- /dev/null +++ b/outputs/2026-05-31/16-33-11/.hydra/config.yaml @@ -0,0 +1,235 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 8 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 700 + customized_validation: true + memory_condition_length: 0 + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + memory_condition_length: 0 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + debug_force_all_streams: false + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + token_patch_size: 2 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 3 + dynamic: + enabled: true + exclude_latest_local_frames: 4 + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + high_quality_fov_threshold: 0.7 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 3 + stage_policy: + noise_bucket_logging: true + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + dit_hidden_size: 1024 + anchor_gate: 1.0 + dynamic_gate: 1.0 + revisit_gate: 1.0 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: DeMemWM + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_full_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_full_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-05-31/16-33-11/.hydra/hydra.yaml b/outputs/2026-05-31/16-33-11/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..1e7e791218d6ed726b9807159f5e84e551825612 --- /dev/null +++ b/outputs/2026-05-31/16-33-11/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_full_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=700 + - +dataset.customized_validation=true + - +dataset.memory_condition_length=0 + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_force_all_streams=false + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=3 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=3 + - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=8 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=3,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_force_all_streams=false,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.exclude_latest_local_frames=4,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=3,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.stage_policy.noise_bucket_logging=true,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.memory_condition_length=0,+dataset.n_frames_valid=700,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_full_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_full_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=8,experiment.validation.val_every_n_step=2000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-05-31/16-33-11 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-05-31/16-33-11/.hydra/overrides.yaml b/outputs/2026-05-31/16-33-11/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..47fd8dfd4320154e59dcf5d378b89f43446db9b2 --- /dev/null +++ b/outputs/2026-05-31/16-33-11/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_full_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=700 +- +dataset.customized_validation=true +- +dataset.memory_condition_length=0 +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_force_all_streams=false +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=3 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=3 +- ++algorithm.dememwm.stage_policy.noise_bucket_logging=true +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=8 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-05-31/16-33-11/main.log b/outputs/2026-05-31/16-33-11/main.log new file mode 100644 index 0000000000000000000000000000000000000000..272f8f57cb3a695f81cdb0a6d94b95a61cb322f2 --- /dev/null +++ b/outputs/2026-05-31/16-33-11/main.log @@ -0,0 +1,39 @@ +[2026-05-31 16:33:58,507][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-05-31 16:33:58,787][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-05-31 16:33:58,787][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-05-31 16:33:58,787][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-05-31 16:33:58,787][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-05-31 16:34:01,148][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-05-31 16:34:02,180][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-05-31 16:34:05,142][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-05-31 16:34:14,613][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-05-31 16:34:20,380][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-05-31 16:34:34,882][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-05-31 16:34:34,883][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-05-31 16:34:36,691][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 199 K +2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 591 K +3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 591 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +6 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +911 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,573.841 Total estimated model params size (MB) +[2026-05-31 16:34:39,689][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-05-31 16:34:39,689][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-03 16:33:00,183][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-03 16:33:00,184][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-03 16:33:01,749][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-03 16:33:01,750][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-01/15-41-01/.hydra/config.yaml b/outputs/2026-06-01/15-41-01/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..524d4d518b4f396f3129bef0c51692c254595dc9 --- /dev/null +++ b/outputs/2026-06-01/15-41-01/.hydra/config.yaml @@ -0,0 +1,234 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 2 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /share_1/users/bonan_ding/worldmem_data/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /share_1/users/bonan_ding/worldmem_data/minecraft/vae_features + n_frames_valid: 1100 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: + - 3 + - 7 + - 11 + - 15 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + debug_force_all_streams: false + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + stage_policy: + noise_bucket_logging: true + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors +vae_path: /share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-01/15-41-01/.hydra/hydra.yaml b/outputs/2026-06-01/15-41-01/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..241213c048d14694dff0605716740d2924303b49 --- /dev/null +++ b/outputs/2026-06-01/15-41-01/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors + - +vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft + - dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=1100 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_force_all_streams=false + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=2 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_force_all_streams=false,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.stage_policy.noise_bucket_logging=true,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=1100,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features,dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=2,experiment.validation.val_every_n_step=2000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-01/15-41-01 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-01/15-41-01/.hydra/overrides.yaml b/outputs/2026-06-01/15-41-01/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5ad2fae2e9cad8bfdc384b11f419b198a35e608d --- /dev/null +++ b/outputs/2026-06-01/15-41-01/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors +- +vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft +- dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=1100 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_force_all_streams=false +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.stage_policy.noise_bucket_logging=true +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=2 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-01/15-41-01/main.log b/outputs/2026-06-01/15-41-01/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/2026-06-01/15-41-06/.hydra/config.yaml b/outputs/2026-06-01/15-41-06/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..524d4d518b4f396f3129bef0c51692c254595dc9 --- /dev/null +++ b/outputs/2026-06-01/15-41-06/.hydra/config.yaml @@ -0,0 +1,234 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 2 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /share_1/users/bonan_ding/worldmem_data/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /share_1/users/bonan_ding/worldmem_data/minecraft/vae_features + n_frames_valid: 1100 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: + - 3 + - 7 + - 11 + - 15 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + debug_force_all_streams: false + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + stage_policy: + noise_bucket_logging: true + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors +vae_path: /share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-01/15-41-06/.hydra/hydra.yaml b/outputs/2026-06-01/15-41-06/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4326c507342c10f1a725d2d6946937ed7382db0e --- /dev/null +++ b/outputs/2026-06-01/15-41-06/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors + - +vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft + - dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=1100 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_force_all_streams=false + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=2 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_force_all_streams=false,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.stage_policy.noise_bucket_logging=true,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=1100,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features,dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=2,experiment.validation.val_every_n_step=2000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-01/15-41-06 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-01/15-41-06/.hydra/overrides.yaml b/outputs/2026-06-01/15-41-06/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5ad2fae2e9cad8bfdc384b11f419b198a35e608d --- /dev/null +++ b/outputs/2026-06-01/15-41-06/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors +- +vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft +- dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=1100 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_force_all_streams=false +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.stage_policy.noise_bucket_logging=true +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=2 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-01/15-41-06/main.log b/outputs/2026-06-01/15-41-06/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/2026-06-01/15-41-07/.hydra/config.yaml b/outputs/2026-06-01/15-41-07/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..524d4d518b4f396f3129bef0c51692c254595dc9 --- /dev/null +++ b/outputs/2026-06-01/15-41-07/.hydra/config.yaml @@ -0,0 +1,234 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 2 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /share_1/users/bonan_ding/worldmem_data/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /share_1/users/bonan_ding/worldmem_data/minecraft/vae_features + n_frames_valid: 1100 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: + - 3 + - 7 + - 11 + - 15 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + debug_force_all_streams: false + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + stage_policy: + noise_bucket_logging: true + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors +vae_path: /share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-01/15-41-07/.hydra/hydra.yaml b/outputs/2026-06-01/15-41-07/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..236d86387dd98ce428f76a20afaea78060333986 --- /dev/null +++ b/outputs/2026-06-01/15-41-07/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors + - +vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft + - dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=1100 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_force_all_streams=false + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=2 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_force_all_streams=false,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.stage_policy.noise_bucket_logging=true,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=1100,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features,dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=2,experiment.validation.val_every_n_step=2000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-01/15-41-07 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-01/15-41-07/.hydra/overrides.yaml b/outputs/2026-06-01/15-41-07/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5ad2fae2e9cad8bfdc384b11f419b198a35e608d --- /dev/null +++ b/outputs/2026-06-01/15-41-07/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors +- +vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft +- dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=1100 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_force_all_streams=false +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.stage_policy.noise_bucket_logging=true +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=2 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-01/15-41-07/main.log b/outputs/2026-06-01/15-41-07/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/2026-06-01/15-41-08/.hydra/config.yaml b/outputs/2026-06-01/15-41-08/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..524d4d518b4f396f3129bef0c51692c254595dc9 --- /dev/null +++ b/outputs/2026-06-01/15-41-08/.hydra/config.yaml @@ -0,0 +1,234 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 2 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /share_1/users/bonan_ding/worldmem_data/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /share_1/users/bonan_ding/worldmem_data/minecraft/vae_features + n_frames_valid: 1100 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: + - 3 + - 7 + - 11 + - 15 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + debug_force_all_streams: false + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + stage_policy: + noise_bucket_logging: true + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors +vae_path: /share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-01/15-41-08/.hydra/hydra.yaml b/outputs/2026-06-01/15-41-08/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..947a3c4e9eecddc25e23cbe9b9ed0a321639fcff --- /dev/null +++ b/outputs/2026-06-01/15-41-08/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors + - +vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft + - dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=1100 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_force_all_streams=false + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=2 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_force_all_streams=false,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.stage_policy.noise_bucket_logging=true,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=1100,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features,dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=2,experiment.validation.val_every_n_step=2000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-01/15-41-08 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-01/15-41-08/.hydra/overrides.yaml b/outputs/2026-06-01/15-41-08/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5ad2fae2e9cad8bfdc384b11f419b198a35e608d --- /dev/null +++ b/outputs/2026-06-01/15-41-08/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors +- +vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft +- dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=1100 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_force_all_streams=false +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.stage_policy.noise_bucket_logging=true +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=2 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-01/15-41-08/main.log b/outputs/2026-06-01/15-41-08/main.log new file mode 100644 index 0000000000000000000000000000000000000000..0514aa6f3467861d38e1830185e8ef48a7b12d9e --- /dev/null +++ b/outputs/2026-06-01/15-41-08/main.log @@ -0,0 +1,5 @@ +[2026-06-01 15:41:49,208][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-01 15:41:49,334][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-01 15:41:49,334][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-01 15:41:49,334][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-01 15:41:49,334][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs diff --git a/outputs/2026-06-03/20-22-54/.hydra/config.yaml b/outputs/2026-06-03/20-22-54/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..bc967412e28848c9107a72c04ac40cbb8f36791b --- /dev/null +++ b/outputs/2026-06-03/20-22-54/.hydra/config.yaml @@ -0,0 +1,234 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 2 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 1100 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: + - 3 + - 7 + - 11 + - 15 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + debug_force_all_streams: false + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + stage_policy: + noise_bucket_logging: true + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-03/20-22-54/.hydra/hydra.yaml b/outputs/2026-06-03/20-22-54/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c8967584d5c9855c45197e0d05b0360aec9c8471 --- /dev/null +++ b/outputs/2026-06-03/20-22-54/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=1100 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_force_all_streams=false + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=2 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_force_all_streams=false,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.stage_policy.noise_bucket_logging=true,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=1100,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=2,experiment.validation.val_every_n_step=2000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-03/20-22-54 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-03/20-22-54/.hydra/overrides.yaml b/outputs/2026-06-03/20-22-54/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..1de347908bc11f34050a2011e470bf96dc9d5be4 --- /dev/null +++ b/outputs/2026-06-03/20-22-54/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=1100 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_force_all_streams=false +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.stage_policy.noise_bucket_logging=true +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=2 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-03/20-22-54/main.log b/outputs/2026-06-03/20-22-54/main.log new file mode 100644 index 0000000000000000000000000000000000000000..f755f313681872bd09cfbac3089458f620a4c049 --- /dev/null +++ b/outputs/2026-06-03/20-22-54/main.log @@ -0,0 +1,47 @@ +[2026-06-03 20:23:36,031][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-03 20:23:36,800][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-03 20:23:36,800][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-03 20:23:36,800][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-03 20:23:36,800][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-03 20:23:38,339][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-03 20:23:39,236][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-03 20:23:42,001][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-03 20:23:42,005][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-03 20:23:42,015][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-03 20:23:42,026][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-03 20:23:48,581][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-03 20:24:03,625][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-03 20:24:03,625][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-03 20:24:03,625][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-03 20:24:03,625][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-03 20:24:05,745][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 631 M +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 887 K +2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 296 K +3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 296 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +6 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +632 M Trainable params +231 M Non-trainable params +864 M Total params +3,456.995 Total estimated model params size (MB) +[2026-06-03 20:24:08,742][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-03 20:24:08,742][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-03 20:24:08,742][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-03 20:24:08,742][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-04 06:21:13,812][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-04 06:21:13,813][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-04 06:21:13,812][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-04 06:21:13,813][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-04 06:21:13,812][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-04 06:21:13,812][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-04 06:21:13,814][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-04 06:21:13,814][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-03/20-22-55/.hydra/config.yaml b/outputs/2026-06-03/20-22-55/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..bc967412e28848c9107a72c04ac40cbb8f36791b --- /dev/null +++ b/outputs/2026-06-03/20-22-55/.hydra/config.yaml @@ -0,0 +1,234 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 2 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 1100 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: + - 3 + - 7 + - 11 + - 15 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + debug_force_all_streams: false + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + stage_policy: + noise_bucket_logging: true + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-03/20-22-55/.hydra/hydra.yaml b/outputs/2026-06-03/20-22-55/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..6a19691cd6bafafb4a5b56a2ce0629703bebfb39 --- /dev/null +++ b/outputs/2026-06-03/20-22-55/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=1100 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_force_all_streams=false + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=2 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_force_all_streams=false,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.stage_policy.noise_bucket_logging=true,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=1100,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=2,experiment.validation.val_every_n_step=2000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-03/20-22-55 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-03/20-22-55/.hydra/overrides.yaml b/outputs/2026-06-03/20-22-55/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..1de347908bc11f34050a2011e470bf96dc9d5be4 --- /dev/null +++ b/outputs/2026-06-03/20-22-55/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=1100 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_force_all_streams=false +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.stage_policy.noise_bucket_logging=true +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=2 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-03/20-22-55/main.log b/outputs/2026-06-03/20-22-55/main.log new file mode 100644 index 0000000000000000000000000000000000000000..3f44cc084f0c750896c8374c9eb97446321660f2 --- /dev/null +++ b/outputs/2026-06-03/20-22-55/main.log @@ -0,0 +1,10 @@ +[2026-06-03 20:23:41,996][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-03 20:23:42,011][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-03 20:24:03,625][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-03 20:24:03,625][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-03 20:24:08,742][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-03 20:24:08,742][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-04 06:21:13,812][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-04 06:21:13,813][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-04 06:21:13,812][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-04 06:21:13,813][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-03/20-22-56/.hydra/config.yaml b/outputs/2026-06-03/20-22-56/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..bc967412e28848c9107a72c04ac40cbb8f36791b --- /dev/null +++ b/outputs/2026-06-03/20-22-56/.hydra/config.yaml @@ -0,0 +1,234 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 2 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 1100 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: + - 3 + - 7 + - 11 + - 15 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + debug_force_all_streams: false + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + stage_policy: + noise_bucket_logging: true + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-03/20-22-56/.hydra/hydra.yaml b/outputs/2026-06-03/20-22-56/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b102173e31cb149ba6b368db51c49d75e65a1abf --- /dev/null +++ b/outputs/2026-06-03/20-22-56/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=1100 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_force_all_streams=false + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=2 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_force_all_streams=false,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.stage_policy.noise_bucket_logging=true,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=1100,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=2,experiment.validation.val_every_n_step=2000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-03/20-22-56 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-03/20-22-56/.hydra/overrides.yaml b/outputs/2026-06-03/20-22-56/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..1de347908bc11f34050a2011e470bf96dc9d5be4 --- /dev/null +++ b/outputs/2026-06-03/20-22-56/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=1100 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_force_all_streams=false +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.stage_policy.noise_bucket_logging=true +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=2 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-03/20-22-56/main.log b/outputs/2026-06-03/20-22-56/main.log new file mode 100644 index 0000000000000000000000000000000000000000..dd2a8c604e8450b593daa9824b78acc23b90952a --- /dev/null +++ b/outputs/2026-06-03/20-22-56/main.log @@ -0,0 +1,10 @@ +[2026-06-03 20:23:42,010][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-03 20:23:42,017][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-03 20:24:03,625][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-03 20:24:03,625][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-03 20:24:08,742][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-03 20:24:08,742][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-04 06:21:13,812][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-04 06:21:13,813][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-04 06:21:13,813][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-04 06:21:13,813][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-04/06-22-27/.hydra/config.yaml b/outputs/2026-06-04/06-22-27/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3640bb3cbb3cb04f1b3f8269c7c813f70d37cc96 --- /dev/null +++ b/outputs/2026-06-04/06-22-27/.hydra/config.yaml @@ -0,0 +1,231 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: + - 3 + - 7 + - 11 + - 15 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-04/06-22-27/.hydra/hydra.yaml b/outputs/2026-06-04/06-22-27/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9728d6e3da029ca14b33ba70bc101a4014d8c851 --- /dev/null +++ b/outputs/2026-06-04/06-22-27/.hydra/hydra.yaml @@ -0,0 +1,230 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-04/06-22-27 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-04/06-22-27/.hydra/overrides.yaml b/outputs/2026-06-04/06-22-27/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f5cfcc49c373d20f7b9d19b1aa4a76e9f3840859 --- /dev/null +++ b/outputs/2026-06-04/06-22-27/.hydra/overrides.yaml @@ -0,0 +1,72 @@ +- +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-04/06-22-27/main.log b/outputs/2026-06-04/06-22-27/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/2026-06-04/06-22-28/.hydra/config.yaml b/outputs/2026-06-04/06-22-28/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3640bb3cbb3cb04f1b3f8269c7c813f70d37cc96 --- /dev/null +++ b/outputs/2026-06-04/06-22-28/.hydra/config.yaml @@ -0,0 +1,231 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: + - 3 + - 7 + - 11 + - 15 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-04/06-22-28/.hydra/hydra.yaml b/outputs/2026-06-04/06-22-28/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d005c001f5cdedc82042c05a287d29358d89749a --- /dev/null +++ b/outputs/2026-06-04/06-22-28/.hydra/hydra.yaml @@ -0,0 +1,230 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-04/06-22-28 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-04/06-22-28/.hydra/overrides.yaml b/outputs/2026-06-04/06-22-28/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f5cfcc49c373d20f7b9d19b1aa4a76e9f3840859 --- /dev/null +++ b/outputs/2026-06-04/06-22-28/.hydra/overrides.yaml @@ -0,0 +1,72 @@ +- +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-04/06-22-28/main.log b/outputs/2026-06-04/06-22-28/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/2026-06-04/06-24-27/.hydra/config.yaml b/outputs/2026-06-04/06-24-27/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8e50746541fe21844337576173118a055f92a686 --- /dev/null +++ b/outputs/2026-06-04/06-24-27/.hydra/config.yaml @@ -0,0 +1,232 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + memory_condition_length: 0 + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + memory_condition_length: 0 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + token_patch_size: 2 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 3 + dynamic: + enabled: true + exclude_latest_local_frames: 4 + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + high_quality_fov_threshold: 0.7 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 3 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + dit_hidden_size: 1024 + anchor_gate: 1.0 + dynamic_gate: 1.0 + revisit_gate: 1.0 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_full_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-04/06-24-27/.hydra/hydra.yaml b/outputs/2026-06-04/06-24-27/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4a64186936bf540d7cda57a16db90faa1560ce8e --- /dev/null +++ b/outputs/2026-06-04/06-24-27/.hydra/hydra.yaml @@ -0,0 +1,230 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.memory_condition_length=0 + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=3 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=3 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=3,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.exclude_latest_local_frames=4,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=3,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.memory_condition_length=0,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_full_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-04/06-24-27 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-04/06-24-27/.hydra/overrides.yaml b/outputs/2026-06-04/06-24-27/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9827984995b3535ee762f6a1370ef381be13c82e --- /dev/null +++ b/outputs/2026-06-04/06-24-27/.hydra/overrides.yaml @@ -0,0 +1,72 @@ +- +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.memory_condition_length=0 +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=3 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=3 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-04/06-24-27/main.log b/outputs/2026-06-04/06-24-27/main.log new file mode 100644 index 0000000000000000000000000000000000000000..9d93adadbd86ace21ba8e1c5fabe75373d22ec88 --- /dev/null +++ b/outputs/2026-06-04/06-24-27/main.log @@ -0,0 +1,77 @@ +[2026-06-04 06:24:58,261][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-04 06:24:58,374][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-04 06:24:58,374][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-04 06:24:58,374][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-04 06:24:58,374][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-04 06:24:58,374][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-04 06:24:58,617][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-04 06:24:58,844][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-04 06:25:01,130][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-04 06:25:01,209][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-04 06:25:01,432][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-04 06:25:01,930][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-04 06:25:02,129][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-04 06:25:02,209][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-04 06:25:03,309][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-04 06:25:07,333][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-04 06:25:20,836][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-04 06:25:20,836][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-04 06:25:20,836][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-04 06:25:20,836][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-04 06:25:20,836][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-04 06:25:20,836][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-04 06:25:20,837][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-04 06:25:21,907][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 199 K +2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 591 K +3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 591 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +6 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +911 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,573.841 Total estimated model params size (MB) +[2026-06-04 06:25:22,815][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-04 06:25:22,815][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-04 06:25:22,815][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-04 06:25:22,815][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-04 06:25:22,815][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-04 06:25:22,815][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-04 06:25:22,815][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-07 06:24:48,015][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-07 06:24:48,015][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-07 06:24:48,016][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-07 06:24:48,015][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-07 06:24:48,016][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-07 06:24:48,015][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-07 06:24:48,016][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-07 06:24:48,016][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-07 06:24:48,015][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-07 06:24:48,016][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-07 06:24:48,016][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-07 06:24:48,015][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-07 06:24:48,016][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-07 06:24:48,016][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-07 06:24:48,021][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-07 06:24:48,021][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-07 06:24:48,021][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-07 06:24:48,021][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-07 06:24:48,021][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-04/06-24-28/.hydra/config.yaml b/outputs/2026-06-04/06-24-28/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8e50746541fe21844337576173118a055f92a686 --- /dev/null +++ b/outputs/2026-06-04/06-24-28/.hydra/config.yaml @@ -0,0 +1,232 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + memory_condition_length: 0 + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + memory_condition_length: 0 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + token_patch_size: 2 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 3 + dynamic: + enabled: true + exclude_latest_local_frames: 4 + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + high_quality_fov_threshold: 0.7 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 3 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + dit_hidden_size: 1024 + anchor_gate: 1.0 + dynamic_gate: 1.0 + revisit_gate: 1.0 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_full_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-04/06-24-28/.hydra/hydra.yaml b/outputs/2026-06-04/06-24-28/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5db771ecded429b21a919779b187c22a9754c67d --- /dev/null +++ b/outputs/2026-06-04/06-24-28/.hydra/hydra.yaml @@ -0,0 +1,230 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.memory_condition_length=0 + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=3 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=3 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=3,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.exclude_latest_local_frames=4,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=3,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.memory_condition_length=0,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_full_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-04/06-24-28 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-04/06-24-28/.hydra/overrides.yaml b/outputs/2026-06-04/06-24-28/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9827984995b3535ee762f6a1370ef381be13c82e --- /dev/null +++ b/outputs/2026-06-04/06-24-28/.hydra/overrides.yaml @@ -0,0 +1,72 @@ +- +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.memory_condition_length=0 +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=3 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=3 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-04/06-24-28/main.log b/outputs/2026-06-04/06-24-28/main.log new file mode 100644 index 0000000000000000000000000000000000000000..0a28a589b244b086011356493a633db24faacc7b --- /dev/null +++ b/outputs/2026-06-04/06-24-28/main.log @@ -0,0 +1,7 @@ +[2026-06-04 06:25:01,236][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-04 06:25:20,836][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-04 06:25:22,816][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-07 06:24:48,015][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-07 06:24:48,015][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-04/06-56-07/.hydra/config.yaml b/outputs/2026-06-04/06-56-07/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a4ac400fccef6e5a5d7e37db8dbefb5a01135211 --- /dev/null +++ b/outputs/2026-06-04/06-56-07/.hydra/config.yaml @@ -0,0 +1,232 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: + - 3 + - 7 + - 11 + - 15 + pose_cond_dim: 5 + log_video: true + memory_condition_length: 0 + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-04/06-56-07/.hydra/hydra.yaml b/outputs/2026-06-04/06-56-07/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..76aef61d23a699b88bde5806bbde2768b034f705 --- /dev/null +++ b/outputs/2026-06-04/06-56-07/.hydra/hydra.yaml @@ -0,0 +1,230 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-04/06-56-07 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-04/06-56-07/.hydra/overrides.yaml b/outputs/2026-06-04/06-56-07/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f5cfcc49c373d20f7b9d19b1aa4a76e9f3840859 --- /dev/null +++ b/outputs/2026-06-04/06-56-07/.hydra/overrides.yaml @@ -0,0 +1,72 @@ +- +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-04/06-56-07/main.log b/outputs/2026-06-04/06-56-07/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/2026-06-04/06-56-08/.hydra/config.yaml b/outputs/2026-06-04/06-56-08/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a4ac400fccef6e5a5d7e37db8dbefb5a01135211 --- /dev/null +++ b/outputs/2026-06-04/06-56-08/.hydra/config.yaml @@ -0,0 +1,232 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: + - 3 + - 7 + - 11 + - 15 + pose_cond_dim: 5 + log_video: true + memory_condition_length: 0 + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-04/06-56-08/.hydra/hydra.yaml b/outputs/2026-06-04/06-56-08/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..bdccbcee233f9548be589ba7ae1bd8dd90840782 --- /dev/null +++ b/outputs/2026-06-04/06-56-08/.hydra/hydra.yaml @@ -0,0 +1,230 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-04/06-56-08 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-04/06-56-08/.hydra/overrides.yaml b/outputs/2026-06-04/06-56-08/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f5cfcc49c373d20f7b9d19b1aa4a76e9f3840859 --- /dev/null +++ b/outputs/2026-06-04/06-56-08/.hydra/overrides.yaml @@ -0,0 +1,72 @@ +- +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-04/06-56-08/main.log b/outputs/2026-06-04/06-56-08/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/2026-06-04/06-56-09/.hydra/config.yaml b/outputs/2026-06-04/06-56-09/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a4ac400fccef6e5a5d7e37db8dbefb5a01135211 --- /dev/null +++ b/outputs/2026-06-04/06-56-09/.hydra/config.yaml @@ -0,0 +1,232 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: + - 3 + - 7 + - 11 + - 15 + pose_cond_dim: 5 + log_video: true + memory_condition_length: 0 + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-04/06-56-09/.hydra/hydra.yaml b/outputs/2026-06-04/06-56-09/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..70af697fb90bc595008a410b8a8033183b5b8cb1 --- /dev/null +++ b/outputs/2026-06-04/06-56-09/.hydra/hydra.yaml @@ -0,0 +1,230 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-04/06-56-09 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-04/06-56-09/.hydra/overrides.yaml b/outputs/2026-06-04/06-56-09/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f5cfcc49c373d20f7b9d19b1aa4a76e9f3840859 --- /dev/null +++ b/outputs/2026-06-04/06-56-09/.hydra/overrides.yaml @@ -0,0 +1,72 @@ +- +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-04/06-56-09/main.log b/outputs/2026-06-04/06-56-09/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/2026-06-05/04-16-56/.hydra/config.yaml b/outputs/2026-06-05/04-16-56/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3640bb3cbb3cb04f1b3f8269c7c813f70d37cc96 --- /dev/null +++ b/outputs/2026-06-05/04-16-56/.hydra/config.yaml @@ -0,0 +1,231 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: + - 3 + - 7 + - 11 + - 15 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-05/04-16-56/.hydra/hydra.yaml b/outputs/2026-06-05/04-16-56/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..768c93711a10688babfde34234b9a1b65faefc2c --- /dev/null +++ b/outputs/2026-06-05/04-16-56/.hydra/hydra.yaml @@ -0,0 +1,230 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-05/04-16-56 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-05/04-16-56/.hydra/overrides.yaml b/outputs/2026-06-05/04-16-56/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f5cfcc49c373d20f7b9d19b1aa4a76e9f3840859 --- /dev/null +++ b/outputs/2026-06-05/04-16-56/.hydra/overrides.yaml @@ -0,0 +1,72 @@ +- +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-05/04-16-56/main.log b/outputs/2026-06-05/04-16-56/main.log new file mode 100644 index 0000000000000000000000000000000000000000..39aa75d8d27fd8e02bf5092d9941485f93727450 --- /dev/null +++ b/outputs/2026-06-05/04-16-56/main.log @@ -0,0 +1,68 @@ +[2026-06-05 04:17:39,878][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-05 04:17:40,248][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-05 04:17:40,249][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-05 04:17:40,249][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-05 04:17:40,249][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-05 04:17:40,249][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-05 04:17:43,881][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-05 04:17:44,806][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-05 04:17:47,641][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-05 04:17:47,641][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-05 04:17:47,641][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-05 04:17:47,641][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-05 04:17:47,641][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-05 04:17:47,641][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-05 04:17:47,647][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-05 04:17:47,654][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-05 04:17:54,247][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-05 04:18:08,625][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-05 04:18:08,625][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-05 04:18:08,625][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-05 04:18:08,625][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-05 04:18:08,625][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-05 04:18:08,626][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-05 04:18:08,625][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-05 04:18:08,626][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-05 04:18:11,315][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 631 M +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 887 K +2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 296 K +3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 296 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +6 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +632 M Trainable params +231 M Non-trainable params +864 M Total params +3,456.995 Total estimated model params size (MB) +[2026-06-05 04:18:13,299][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-05 04:18:13,299][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-05 04:18:13,299][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-05 04:18:13,299][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-05 04:18:13,299][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-05 04:18:13,299][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-05 04:18:13,299][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-05 04:18:13,299][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-08 04:16:57,057][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-08 04:16:57,057][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-08 04:16:57,057][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-08 04:16:57,058][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-08 04:16:57,057][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-08 04:16:57,057][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-08 04:16:57,058][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-08 04:16:57,057][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-08 04:16:57,058][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-08 04:16:57,057][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-08 04:16:57,058][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-08 04:16:57,058][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-08 04:16:57,057][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-08 04:16:57,059][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-08 04:16:57,057][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-08 04:16:57,059][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-07/16-35-26/.hydra/config.yaml b/outputs/2026-06-07/16-35-26/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8e50746541fe21844337576173118a055f92a686 --- /dev/null +++ b/outputs/2026-06-07/16-35-26/.hydra/config.yaml @@ -0,0 +1,232 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + memory_condition_length: 0 + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + memory_condition_length: 0 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + token_patch_size: 2 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 3 + dynamic: + enabled: true + exclude_latest_local_frames: 4 + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + high_quality_fov_threshold: 0.7 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 3 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + dit_hidden_size: 1024 + anchor_gate: 1.0 + dynamic_gate: 1.0 + revisit_gate: 1.0 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_full_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-07/16-35-26/.hydra/hydra.yaml b/outputs/2026-06-07/16-35-26/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..81a82156d72c3752fbe5c603233054b88bea8007 --- /dev/null +++ b/outputs/2026-06-07/16-35-26/.hydra/hydra.yaml @@ -0,0 +1,230 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.memory_condition_length=0 + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=3 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=3 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=3,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.exclude_latest_local_frames=4,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=3,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.memory_condition_length=0,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_full_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-07/16-35-26 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-07/16-35-26/.hydra/overrides.yaml b/outputs/2026-06-07/16-35-26/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9827984995b3535ee762f6a1370ef381be13c82e --- /dev/null +++ b/outputs/2026-06-07/16-35-26/.hydra/overrides.yaml @@ -0,0 +1,72 @@ +- +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.memory_condition_length=0 +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=3 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=3 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-07/16-35-26/main.log b/outputs/2026-06-07/16-35-26/main.log new file mode 100644 index 0000000000000000000000000000000000000000..30116287f976764ee9144b1593522dc290d1fb31 --- /dev/null +++ b/outputs/2026-06-07/16-35-26/main.log @@ -0,0 +1,41 @@ +[2026-06-07 16:36:13,566][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-07 16:36:13,664][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-07 16:36:13,664][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-07 16:36:13,664][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-07 16:36:13,664][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-07 16:36:13,665][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-07 16:36:14,033][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-07 16:36:14,933][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-07 16:36:15,013][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-07 16:36:15,983][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-07 16:36:22,895][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-07 16:36:36,654][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/checkpoints/last.ckpt +[2026-06-07 16:36:54,144][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-07 16:36:54,798][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-07 16:36:54,867][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-07 16:36:54,942][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-07 16:36:55,760][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 199 K +2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 591 K +3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 591 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +6 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +911 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,573.841 Total estimated model params size (MB) +[2026-06-07 16:36:56,564][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/checkpoints/last.ckpt +[2026-06-07 16:36:59,258][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-07 16:36:59,258][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-07 16:36:59,258][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-07 16:36:59,258][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-10 05:47:16,693][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=350000` reached. diff --git a/outputs/2026-06-07/16-35-27/.hydra/config.yaml b/outputs/2026-06-07/16-35-27/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8e50746541fe21844337576173118a055f92a686 --- /dev/null +++ b/outputs/2026-06-07/16-35-27/.hydra/config.yaml @@ -0,0 +1,232 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + memory_condition_length: 0 + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + memory_condition_length: 0 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + token_patch_size: 2 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 3 + dynamic: + enabled: true + exclude_latest_local_frames: 4 + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + high_quality_fov_threshold: 0.7 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 3 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + dit_hidden_size: 1024 + anchor_gate: 1.0 + dynamic_gate: 1.0 + revisit_gate: 1.0 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_full_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-07/16-35-27/.hydra/hydra.yaml b/outputs/2026-06-07/16-35-27/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a09cc5cf14121c8febd7d6fac46b73f7e298ab85 --- /dev/null +++ b/outputs/2026-06-07/16-35-27/.hydra/hydra.yaml @@ -0,0 +1,230 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.memory_condition_length=0 + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=3 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=3 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=3,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.exclude_latest_local_frames=4,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=3,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.memory_condition_length=0,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_full_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-07/16-35-27 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-07/16-35-27/.hydra/overrides.yaml b/outputs/2026-06-07/16-35-27/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9827984995b3535ee762f6a1370ef381be13c82e --- /dev/null +++ b/outputs/2026-06-07/16-35-27/.hydra/overrides.yaml @@ -0,0 +1,72 @@ +- +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.memory_condition_length=0 +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=3 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=3 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-07/16-35-27/main.log b/outputs/2026-06-07/16-35-27/main.log new file mode 100644 index 0000000000000000000000000000000000000000..bfa392e49802ad5dbddbf3928a9965c464f1e981 --- /dev/null +++ b/outputs/2026-06-07/16-35-27/main.log @@ -0,0 +1,6 @@ +[2026-06-07 16:36:13,993][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-07 16:36:14,015][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-07 16:36:54,140][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-07 16:36:54,886][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-07 16:36:59,258][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-07 16:36:59,258][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-07/16-35-28/.hydra/config.yaml b/outputs/2026-06-07/16-35-28/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8e50746541fe21844337576173118a055f92a686 --- /dev/null +++ b/outputs/2026-06-07/16-35-28/.hydra/config.yaml @@ -0,0 +1,232 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + memory_condition_length: 0 + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + memory_condition_length: 0 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + token_patch_size: 2 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 3 + dynamic: + enabled: true + exclude_latest_local_frames: 4 + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + high_quality_fov_threshold: 0.7 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 3 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + dit_hidden_size: 1024 + anchor_gate: 1.0 + dynamic_gate: 1.0 + revisit_gate: 1.0 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_full_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-07/16-35-28/.hydra/hydra.yaml b/outputs/2026-06-07/16-35-28/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..aa5fb6c56cc4651d1ebe01b7967edfcbff87e1c8 --- /dev/null +++ b/outputs/2026-06-07/16-35-28/.hydra/hydra.yaml @@ -0,0 +1,230 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.memory_condition_length=0 + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=3 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=3 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=3,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.exclude_latest_local_frames=4,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=3,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.memory_condition_length=0,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_full_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-07/16-35-28 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-07/16-35-28/.hydra/overrides.yaml b/outputs/2026-06-07/16-35-28/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9827984995b3535ee762f6a1370ef381be13c82e --- /dev/null +++ b/outputs/2026-06-07/16-35-28/.hydra/overrides.yaml @@ -0,0 +1,72 @@ +- +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.memory_condition_length=0 +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=3 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=3 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-07/16-35-28/main.log b/outputs/2026-06-07/16-35-28/main.log new file mode 100644 index 0000000000000000000000000000000000000000..b17c008e8543df07bf454a62108560684f4e8f94 --- /dev/null +++ b/outputs/2026-06-07/16-35-28/main.log @@ -0,0 +1,6 @@ +[2026-06-07 16:36:14,069][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-07 16:36:14,085][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-07 16:36:54,051][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-07 16:36:54,868][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-07 16:36:59,258][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-07 16:36:59,258][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-07/21-26-49/.hydra/config.yaml b/outputs/2026-06-07/21-26-49/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..383e38a44d3cc4e0ab79c0c959f3bc58ccd515ee --- /dev/null +++ b/outputs/2026-06-07/21-26-49/.hydra/config.yaml @@ -0,0 +1,227 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 1100 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 1024 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_full_hidden1024_all_layers_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_full_hidden1024_all_layers_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-07/21-26-49/.hydra/hydra.yaml b/outputs/2026-06-07/21-26-49/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a13ec33a69ac59d582598dc9fe88637e82fa108e --- /dev/null +++ b/outputs/2026-06-07/21-26-49/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_full_hidden1024_all_layers_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_full_hidden1024_all_layers_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=1100 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - algorithm.memory_cross_attn_layers=null + - ++algorithm.dememwm.injection.memory_hidden_size=1024 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.injection.memory_hidden_size=1024,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=1100,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_full_hidden1024_all_layers_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_full_hidden1024_all_layers_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.memory_cross_attn_layers=null,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-07/21-26-49 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-07/21-26-49/.hydra/overrides.yaml b/outputs/2026-06-07/21-26-49/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fd80489eda9c1b11b112f3919f467eb50f635c86 --- /dev/null +++ b/outputs/2026-06-07/21-26-49/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_separate_memory_full_hidden1024_all_layers_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_full_hidden1024_all_layers_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=1100 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- algorithm.memory_cross_attn_layers=null +- ++algorithm.dememwm.injection.memory_hidden_size=1024 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-07/21-26-49/main.log b/outputs/2026-06-07/21-26-49/main.log new file mode 100644 index 0000000000000000000000000000000000000000..817132fc528bebf40fed79b60c28215472161e42 --- /dev/null +++ b/outputs/2026-06-07/21-26-49/main.log @@ -0,0 +1,43 @@ +[2026-06-07 21:27:32,831][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-07 21:27:33,847][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-07 21:27:33,847][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-07 21:27:33,847][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-07 21:27:33,847][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-07 21:27:33,848][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-07 21:27:36,265][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-07 21:27:37,179][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-07 21:27:39,857][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-07 21:27:39,859][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-07 21:27:39,866][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-07 21:27:46,350][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-07 21:28:00,514][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-07 21:28:00,514][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-07 21:28:00,514][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-07 21:28:01,607][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 994 M +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 1.8 M +2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 591 K +3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 591 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +6 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +997 M Trainable params +231 M Non-trainable params +1.2 B Total params +4,915.943 Total estimated model params size (MB) +[2026-06-07 21:28:05,410][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-07 21:28:05,410][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-07 21:28:05,410][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-10 21:26:49,109][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-10 21:26:49,109][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-10 21:26:49,110][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-10 21:26:49,110][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-10 21:26:49,110][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-10 21:26:49,111][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-07/21-26-50/.hydra/config.yaml b/outputs/2026-06-07/21-26-50/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..383e38a44d3cc4e0ab79c0c959f3bc58ccd515ee --- /dev/null +++ b/outputs/2026-06-07/21-26-50/.hydra/config.yaml @@ -0,0 +1,227 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 1100 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 1024 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_full_hidden1024_all_layers_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_full_hidden1024_all_layers_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-07/21-26-50/.hydra/hydra.yaml b/outputs/2026-06-07/21-26-50/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..10d01f33aad1dc3f65878fa482cbb5660b84b8ac --- /dev/null +++ b/outputs/2026-06-07/21-26-50/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_full_hidden1024_all_layers_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_full_hidden1024_all_layers_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=1100 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - algorithm.memory_cross_attn_layers=null + - ++algorithm.dememwm.injection.memory_hidden_size=1024 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.injection.memory_hidden_size=1024,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=1100,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_full_hidden1024_all_layers_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_full_hidden1024_all_layers_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.memory_cross_attn_layers=null,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-07/21-26-50 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-07/21-26-50/.hydra/overrides.yaml b/outputs/2026-06-07/21-26-50/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fd80489eda9c1b11b112f3919f467eb50f635c86 --- /dev/null +++ b/outputs/2026-06-07/21-26-50/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_separate_memory_full_hidden1024_all_layers_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_full_hidden1024_all_layers_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=1100 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- algorithm.memory_cross_attn_layers=null +- ++algorithm.dememwm.injection.memory_hidden_size=1024 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-07/21-26-50/main.log b/outputs/2026-06-07/21-26-50/main.log new file mode 100644 index 0000000000000000000000000000000000000000..2dc8e7ef9c5225bce71131954893bf79d870bfb4 --- /dev/null +++ b/outputs/2026-06-07/21-26-50/main.log @@ -0,0 +1,25 @@ +[2026-06-07 21:27:39,859][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-07 21:27:39,859][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-07 21:27:39,859][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-07 21:27:39,863][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-07 21:27:39,866][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-07 21:28:00,514][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-07 21:28:00,514][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-07 21:28:00,514][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-07 21:28:00,514][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-07 21:28:00,514][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-07 21:28:05,410][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-07 21:28:05,410][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-07 21:28:05,410][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-07 21:28:05,410][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-07 21:28:05,410][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-10 21:26:49,110][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-10 21:26:49,110][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-10 21:26:49,110][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-10 21:26:49,110][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-10 21:26:49,111][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-10 21:26:49,111][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-10 21:26:49,111][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-10 21:26:49,111][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-10 21:26:49,110][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-10 21:26:49,111][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-10/18-59-16/.hydra/config.yaml b/outputs/2026-06-10/18-59-16/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..147c36b0f4d0828be9a12297959c5d927b0c80fa --- /dev/null +++ b/outputs/2026-06-10/18-59-16/.hydra/config.yaml @@ -0,0 +1,233 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + memory_condition_length: 0 + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + memory_condition_length: 0 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + token_patch_size: 2 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 3 + dynamic: + enabled: true + exclude_latest_local_frames: 4 + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + revisit: + enabled: true + learn_gate: false + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + high_quality_fov_threshold: 0.7 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 3 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + dit_hidden_size: 1024 + anchor_gate: 1.0 + dynamic_gate: 1.0 + revisit_gate: 1.0 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-10/18-59-16/.hydra/hydra.yaml b/outputs/2026-06-10/18-59-16/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e27f2e3cb518e7f3445395eb78ec2cc5331d6f28 --- /dev/null +++ b/outputs/2026-06-10/18-59-16/.hydra/hydra.yaml @@ -0,0 +1,231 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.memory_condition_length=0 + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=3 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.learn_gate=false + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=3 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=3,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.exclude_latest_local_frames=4,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=3,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.learn_gate=false,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.memory_condition_length=0,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-10/18-59-16 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-10/18-59-16/.hydra/overrides.yaml b/outputs/2026-06-10/18-59-16/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..65ff4ced4ac2e99c0a14f6cfa4b22f1224db9db3 --- /dev/null +++ b/outputs/2026-06-10/18-59-16/.hydra/overrides.yaml @@ -0,0 +1,73 @@ +- +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.memory_condition_length=0 +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=3 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.learn_gate=false +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=3 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-10/18-59-16/main.log b/outputs/2026-06-10/18-59-16/main.log new file mode 100644 index 0000000000000000000000000000000000000000..6bb750a03107f6b3592ebb33d8985cc5aaf7c96c --- /dev/null +++ b/outputs/2026-06-10/18-59-16/main.log @@ -0,0 +1,15 @@ +[2026-06-10 19:00:07,918][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-10 19:00:07,959][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-10 19:00:07,964][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-10 19:00:29,541][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-10 19:00:29,541][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-10 19:00:29,541][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-10 19:00:36,026][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-10 19:00:36,026][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-10 19:00:36,026][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-13 18:59:31,413][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-13 18:59:31,413][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-13 18:59:31,414][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-13 18:59:31,414][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-13 18:59:31,413][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-13 18:59:31,414][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-10/18-59-17/.hydra/config.yaml b/outputs/2026-06-10/18-59-17/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..147c36b0f4d0828be9a12297959c5d927b0c80fa --- /dev/null +++ b/outputs/2026-06-10/18-59-17/.hydra/config.yaml @@ -0,0 +1,233 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + memory_condition_length: 0 + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + memory_condition_length: 0 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + token_patch_size: 2 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 3 + dynamic: + enabled: true + exclude_latest_local_frames: 4 + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + revisit: + enabled: true + learn_gate: false + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + high_quality_fov_threshold: 0.7 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 3 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + dit_hidden_size: 1024 + anchor_gate: 1.0 + dynamic_gate: 1.0 + revisit_gate: 1.0 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-10/18-59-17/.hydra/hydra.yaml b/outputs/2026-06-10/18-59-17/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7b18c3d3b8f0ea108de3d68602cdeb2a49debad1 --- /dev/null +++ b/outputs/2026-06-10/18-59-17/.hydra/hydra.yaml @@ -0,0 +1,231 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.memory_condition_length=0 + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=3 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.learn_gate=false + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=3 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=3,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.exclude_latest_local_frames=4,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=3,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.learn_gate=false,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.memory_condition_length=0,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-10/18-59-17 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-10/18-59-17/.hydra/overrides.yaml b/outputs/2026-06-10/18-59-17/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..65ff4ced4ac2e99c0a14f6cfa4b22f1224db9db3 --- /dev/null +++ b/outputs/2026-06-10/18-59-17/.hydra/overrides.yaml @@ -0,0 +1,73 @@ +- +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.memory_condition_length=0 +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=3 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.learn_gate=false +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=3 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-10/18-59-17/main.log b/outputs/2026-06-10/18-59-17/main.log new file mode 100644 index 0000000000000000000000000000000000000000..30968524713c99b578fe767957a32a7b637bc90c --- /dev/null +++ b/outputs/2026-06-10/18-59-17/main.log @@ -0,0 +1,15 @@ +[2026-06-10 19:00:07,944][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-10 19:00:07,946][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-10 19:00:07,954][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-10 19:00:29,541][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-10 19:00:29,541][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-10 19:00:29,541][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-10 19:00:36,026][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-10 19:00:36,026][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-10 19:00:36,026][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-13 18:59:31,412][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-13 18:59:31,413][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-13 18:59:31,413][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-13 18:59:31,414][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-13 18:59:31,414][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-13 18:59:31,415][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-10/18-59-18/.hydra/config.yaml b/outputs/2026-06-10/18-59-18/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..147c36b0f4d0828be9a12297959c5d927b0c80fa --- /dev/null +++ b/outputs/2026-06-10/18-59-18/.hydra/config.yaml @@ -0,0 +1,233 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + memory_condition_length: 0 + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + memory_condition_length: 0 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + token_patch_size: 2 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 3 + dynamic: + enabled: true + exclude_latest_local_frames: 4 + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + revisit: + enabled: true + learn_gate: false + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + high_quality_fov_threshold: 0.7 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 3 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + dit_hidden_size: 1024 + anchor_gate: 1.0 + dynamic_gate: 1.0 + revisit_gate: 1.0 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-10/18-59-18/.hydra/hydra.yaml b/outputs/2026-06-10/18-59-18/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..94cd061e40d7bc335ae3890f9146bf829f336b6e --- /dev/null +++ b/outputs/2026-06-10/18-59-18/.hydra/hydra.yaml @@ -0,0 +1,231 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.memory_condition_length=0 + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=3 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.learn_gate=false + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=3 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=3,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.exclude_latest_local_frames=4,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=3,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.learn_gate=false,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.memory_condition_length=0,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-10/18-59-18 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-10/18-59-18/.hydra/overrides.yaml b/outputs/2026-06-10/18-59-18/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..65ff4ced4ac2e99c0a14f6cfa4b22f1224db9db3 --- /dev/null +++ b/outputs/2026-06-10/18-59-18/.hydra/overrides.yaml @@ -0,0 +1,73 @@ +- +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k_gate_off/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.memory_condition_length=0 +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=3 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4 +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.learn_gate=false +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=3 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-10/18-59-18/main.log b/outputs/2026-06-10/18-59-18/main.log new file mode 100644 index 0000000000000000000000000000000000000000..ffcb3ae7933baf3d09563a85455e78aa86dd1052 --- /dev/null +++ b/outputs/2026-06-10/18-59-18/main.log @@ -0,0 +1,38 @@ +[2026-06-10 19:00:00,071][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-10 19:00:00,241][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-10 19:00:00,241][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-10 19:00:00,241][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-10 19:00:00,241][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-10 19:00:00,242][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-10 19:00:04,346][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-10 19:00:05,350][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-10 19:00:07,918][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-10 19:00:07,935][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-10 19:00:14,848][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-10 19:00:29,541][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-10 19:00:29,541][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-10 19:00:30,556][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 199 K +2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 591 K +3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 591 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +6 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +911 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,573.841 Total estimated model params size (MB) +[2026-06-10 19:00:36,026][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-10 19:00:36,026][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-13 18:59:31,413][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-13 18:59:31,414][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-13 18:59:31,415][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-13 18:59:31,415][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/00-47-56/.hydra/config.yaml b/outputs/2026-06-12/00-47-56/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..64fffe9c620215d1cab65a6f5f95ac635e29032a --- /dev/null +++ b/outputs/2026-06-12/00-47-56/.hydra/config.yaml @@ -0,0 +1,262 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 750000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 10000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 10000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + training_sampling: + revisit_biased_prob: 0.3 + min_gap: 32 + candidate_starts: 8 + train_negatives: + enabled: true + prob: 0.1 + start_step: 20000 + branches: + - wrong_pose + - time_shuffle + - source_matched_random + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: none + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 20000 + ramp_steps: 20000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + snr_shifted_memory: + enabled: true + mode: step_offset + offset_steps: 1 + apply_anchor: true + apply_dynamic: false + apply_revisit: true + noise_per_target: true + mark_as_generated: false + revisit_metric: + enabled: true + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + injection: + memory_hidden_size: 512 + pose_aware: true + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_gate_none_snr_shift_plucker_mem_8a100_bs4_global32_750k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_plucker_mem_8a100_bs4_global32_750k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/00-47-56/.hydra/hydra.yaml b/outputs/2026-06-12/00-47-56/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a96e636112690ae0aab2b0cbb4fdee97873e5955 --- /dev/null +++ b/outputs/2026-06-12/00-47-56/.hydra/hydra.yaml @@ -0,0 +1,249 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_gate_none_snr_shift_plucker_mem_8a100_bs4_global32_750k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_plucker_mem_8a100_bs4_global32_750k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.injection.pose_aware=true + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=20000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.snr_shifted_memory.enabled=true + - ++algorithm.dememwm.snr_shifted_memory.mode=step_offset + - ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 + - ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true + - ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false + - ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true + - ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true + - ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false + - ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 + - ++algorithm.dememwm.training_sampling.min_gap=32 + - ++algorithm.dememwm.train_negatives.enabled=true + - ++algorithm.dememwm.train_negatives.prob=0.10 + - ++algorithm.dememwm.train_negatives.start_step=20000 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=none + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=10000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=4 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=10000 + - experiment.training.max_steps=750000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=10000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=20000,++algorithm.dememwm.generated_history_proxy.start_step=20000,++algorithm.dememwm.injection.pose_aware=true,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=none,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.snr_shifted_memory.apply_anchor=true,++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false,++algorithm.dememwm.snr_shifted_memory.apply_revisit=true,++algorithm.dememwm.snr_shifted_memory.enabled=true,++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false,++algorithm.dememwm.snr_shifted_memory.mode=step_offset,++algorithm.dememwm.snr_shifted_memory.noise_per_target=true,++algorithm.dememwm.snr_shifted_memory.offset_steps=1,++algorithm.dememwm.train_negatives.enabled=true,++algorithm.dememwm.train_negatives.prob=0.10,++algorithm.dememwm.train_negatives.start_step=20000,++algorithm.dememwm.training_sampling.min_gap=32,++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_gate_none_snr_shift_plucker_mem_8a100_bs4_global32_750k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_plucker_mem_8a100_bs4_global32_750k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=4,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=750000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=10000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/00-47-56 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/00-47-56/.hydra/overrides.yaml b/outputs/2026-06-12/00-47-56/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f4f2bf0676c18a5d49b9d2724af9d00ea079c01f --- /dev/null +++ b/outputs/2026-06-12/00-47-56/.hydra/overrides.yaml @@ -0,0 +1,91 @@ +- +name=train_dememwm_sep_full_seq_gate_none_snr_shift_plucker_mem_8a100_bs4_global32_750k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_plucker_mem_8a100_bs4_global32_750k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.injection.pose_aware=true +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=20000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.snr_shifted_memory.enabled=true +- ++algorithm.dememwm.snr_shifted_memory.mode=step_offset +- ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 +- ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true +- ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false +- ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true +- ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true +- ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false +- ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 +- ++algorithm.dememwm.training_sampling.min_gap=32 +- ++algorithm.dememwm.train_negatives.enabled=true +- ++algorithm.dememwm.train_negatives.prob=0.10 +- ++algorithm.dememwm.train_negatives.start_step=20000 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=none +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=10000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=4 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=10000 +- experiment.training.max_steps=750000 diff --git a/outputs/2026-06-12/00-47-56/main.log b/outputs/2026-06-12/00-47-56/main.log new file mode 100644 index 0000000000000000000000000000000000000000..25dd9d818a2d38c2ef03f781679ab947afaac5c0 --- /dev/null +++ b/outputs/2026-06-12/00-47-56/main.log @@ -0,0 +1,92 @@ +[2026-06-12 00:48:44,382][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-12 00:48:44,549][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-12 00:48:44,549][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-12 00:48:44,549][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-12 00:48:44,549][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-12 00:48:44,549][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-12 00:48:47,035][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 00:48:48,020][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 00:48:50,539][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-12 00:48:50,540][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-12 00:48:50,541][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-12 00:48:50,569][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-12 00:48:50,576][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-12 00:48:57,704][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-12 00:49:14,228][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:14,228][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:14,228][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:14,228][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:14,228][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:15,948][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 834 M +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 887 K +2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 296 K +3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 296 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | dememwm_memory_pose_embed | Sequential | 266 K +6 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +7 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +836 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,273.565 Total estimated model params size (MB) +[2026-06-12 00:49:18,953][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:18,953][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:18,953][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:18,953][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:18,954][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 03:47:19,003][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_pairs', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,013][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_revisit_frames', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,013][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,013][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/non_revisit_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,014][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,014][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,014][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_minus_non_revisit_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,015][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,015][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,015][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_psnr_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,015][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,016][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,016][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,016][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_psnr_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,017][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,017][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,017][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,017][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_psnr_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:47:19,018][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-15 00:47:49,312][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-15 00:47:49,313][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:47:49,312][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-15 00:47:49,313][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:47:49,313][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-15 00:47:49,314][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:47:49,313][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-15 00:47:49,314][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:47:49,312][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-15 00:47:49,315][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/00-47-57/.hydra/config.yaml b/outputs/2026-06-12/00-47-57/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..64fffe9c620215d1cab65a6f5f95ac635e29032a --- /dev/null +++ b/outputs/2026-06-12/00-47-57/.hydra/config.yaml @@ -0,0 +1,262 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 750000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 10000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 10000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + training_sampling: + revisit_biased_prob: 0.3 + min_gap: 32 + candidate_starts: 8 + train_negatives: + enabled: true + prob: 0.1 + start_step: 20000 + branches: + - wrong_pose + - time_shuffle + - source_matched_random + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: none + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 20000 + ramp_steps: 20000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + snr_shifted_memory: + enabled: true + mode: step_offset + offset_steps: 1 + apply_anchor: true + apply_dynamic: false + apply_revisit: true + noise_per_target: true + mark_as_generated: false + revisit_metric: + enabled: true + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + injection: + memory_hidden_size: 512 + pose_aware: true + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_gate_none_snr_shift_plucker_mem_8a100_bs4_global32_750k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_plucker_mem_8a100_bs4_global32_750k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/00-47-57/.hydra/hydra.yaml b/outputs/2026-06-12/00-47-57/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a6d19d055ae57f3903a601e2b5a69ad99971c706 --- /dev/null +++ b/outputs/2026-06-12/00-47-57/.hydra/hydra.yaml @@ -0,0 +1,249 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_gate_none_snr_shift_plucker_mem_8a100_bs4_global32_750k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_plucker_mem_8a100_bs4_global32_750k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.injection.pose_aware=true + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=20000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.snr_shifted_memory.enabled=true + - ++algorithm.dememwm.snr_shifted_memory.mode=step_offset + - ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 + - ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true + - ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false + - ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true + - ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true + - ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false + - ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 + - ++algorithm.dememwm.training_sampling.min_gap=32 + - ++algorithm.dememwm.train_negatives.enabled=true + - ++algorithm.dememwm.train_negatives.prob=0.10 + - ++algorithm.dememwm.train_negatives.start_step=20000 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=none + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=10000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=4 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=10000 + - experiment.training.max_steps=750000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=10000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=20000,++algorithm.dememwm.generated_history_proxy.start_step=20000,++algorithm.dememwm.injection.pose_aware=true,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=none,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.snr_shifted_memory.apply_anchor=true,++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false,++algorithm.dememwm.snr_shifted_memory.apply_revisit=true,++algorithm.dememwm.snr_shifted_memory.enabled=true,++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false,++algorithm.dememwm.snr_shifted_memory.mode=step_offset,++algorithm.dememwm.snr_shifted_memory.noise_per_target=true,++algorithm.dememwm.snr_shifted_memory.offset_steps=1,++algorithm.dememwm.train_negatives.enabled=true,++algorithm.dememwm.train_negatives.prob=0.10,++algorithm.dememwm.train_negatives.start_step=20000,++algorithm.dememwm.training_sampling.min_gap=32,++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_gate_none_snr_shift_plucker_mem_8a100_bs4_global32_750k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_plucker_mem_8a100_bs4_global32_750k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=4,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=750000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=10000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/00-47-57 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/00-47-57/.hydra/overrides.yaml b/outputs/2026-06-12/00-47-57/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f4f2bf0676c18a5d49b9d2724af9d00ea079c01f --- /dev/null +++ b/outputs/2026-06-12/00-47-57/.hydra/overrides.yaml @@ -0,0 +1,91 @@ +- +name=train_dememwm_sep_full_seq_gate_none_snr_shift_plucker_mem_8a100_bs4_global32_750k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_plucker_mem_8a100_bs4_global32_750k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.injection.pose_aware=true +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=20000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.snr_shifted_memory.enabled=true +- ++algorithm.dememwm.snr_shifted_memory.mode=step_offset +- ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 +- ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true +- ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false +- ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true +- ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true +- ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false +- ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 +- ++algorithm.dememwm.training_sampling.min_gap=32 +- ++algorithm.dememwm.train_negatives.enabled=true +- ++algorithm.dememwm.train_negatives.prob=0.10 +- ++algorithm.dememwm.train_negatives.start_step=20000 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=none +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=10000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=4 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=10000 +- experiment.training.max_steps=750000 diff --git a/outputs/2026-06-12/00-47-57/main.log b/outputs/2026-06-12/00-47-57/main.log new file mode 100644 index 0000000000000000000000000000000000000000..3f8966bab1404718bbc2fc08332e93fcfc7dd3f4 --- /dev/null +++ b/outputs/2026-06-12/00-47-57/main.log @@ -0,0 +1,15 @@ +[2026-06-12 00:48:50,540][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-12 00:48:50,540][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-12 00:48:50,540][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-12 00:49:14,228][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:14,228][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:14,228][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:18,953][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:18,953][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:18,953][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 00:47:49,311][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-15 00:47:49,313][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:47:49,313][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-15 00:47:49,313][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:47:49,312][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-15 00:47:49,315][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/00-48-24/.hydra/config.yaml b/outputs/2026-06-12/00-48-24/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0d9cad3f5777e026cecb2ef35be0972cd5179996 --- /dev/null +++ b/outputs/2026-06-12/00-48-24/.hydra/config.yaml @@ -0,0 +1,262 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 750000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 10000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 10000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + training_sampling: + revisit_biased_prob: 0.3 + min_gap: 32 + candidate_starts: 8 + train_negatives: + enabled: true + prob: 0.1 + start_step: 20000 + branches: + - wrong_pose + - time_shuffle + - source_matched_random + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: none + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: dit_patch + downsample_ratio: 1 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 20000 + ramp_steps: 20000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + snr_shifted_memory: + enabled: true + mode: step_offset + offset_steps: 1 + apply_anchor: true + apply_dynamic: false + apply_revisit: true + noise_per_target: true + mark_as_generated: false + revisit_metric: + enabled: true + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + injection: + memory_hidden_size: 512 + pose_aware: true + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_gate_none_snr_shift_rev1_8a100_bs4_global32_750k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_rev1_8a100_bs4_global32_750k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/00-48-24/.hydra/hydra.yaml b/outputs/2026-06-12/00-48-24/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4e157f91d8cb55e4abf17577f47dec6de52d9b35 --- /dev/null +++ b/outputs/2026-06-12/00-48-24/.hydra/hydra.yaml @@ -0,0 +1,249 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_gate_none_snr_shift_rev1_8a100_bs4_global32_750k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_rev1_8a100_bs4_global32_750k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.injection.pose_aware=true + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=20000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.snr_shifted_memory.enabled=true + - ++algorithm.dememwm.snr_shifted_memory.mode=step_offset + - ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 + - ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true + - ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false + - ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true + - ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true + - ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false + - ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 + - ++algorithm.dememwm.training_sampling.min_gap=32 + - ++algorithm.dememwm.train_negatives.enabled=true + - ++algorithm.dememwm.train_negatives.prob=0.10 + - ++algorithm.dememwm.train_negatives.start_step=20000 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=none + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=dit_patch + - ++algorithm.dememwm.revisit.compress.downsample_ratio=1 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=10000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=4 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=10000 + - experiment.training.max_steps=750000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=10000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=20000,++algorithm.dememwm.generated_history_proxy.start_step=20000,++algorithm.dememwm.injection.pose_aware=true,++algorithm.dememwm.revisit.compress.downsample_ratio=1,++algorithm.dememwm.revisit.compress.tokenization=dit_patch,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=none,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.snr_shifted_memory.apply_anchor=true,++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false,++algorithm.dememwm.snr_shifted_memory.apply_revisit=true,++algorithm.dememwm.snr_shifted_memory.enabled=true,++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false,++algorithm.dememwm.snr_shifted_memory.mode=step_offset,++algorithm.dememwm.snr_shifted_memory.noise_per_target=true,++algorithm.dememwm.snr_shifted_memory.offset_steps=1,++algorithm.dememwm.train_negatives.enabled=true,++algorithm.dememwm.train_negatives.prob=0.10,++algorithm.dememwm.train_negatives.start_step=20000,++algorithm.dememwm.training_sampling.min_gap=32,++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_gate_none_snr_shift_rev1_8a100_bs4_global32_750k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_rev1_8a100_bs4_global32_750k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=4,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=750000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=10000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/00-48-24 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/00-48-24/.hydra/overrides.yaml b/outputs/2026-06-12/00-48-24/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3fa71fb829a9374240e37cf84ade058f611e3b71 --- /dev/null +++ b/outputs/2026-06-12/00-48-24/.hydra/overrides.yaml @@ -0,0 +1,91 @@ +- +name=train_dememwm_sep_full_seq_gate_none_snr_shift_rev1_8a100_bs4_global32_750k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_rev1_8a100_bs4_global32_750k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.injection.pose_aware=true +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=20000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.snr_shifted_memory.enabled=true +- ++algorithm.dememwm.snr_shifted_memory.mode=step_offset +- ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 +- ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true +- ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false +- ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true +- ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true +- ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false +- ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 +- ++algorithm.dememwm.training_sampling.min_gap=32 +- ++algorithm.dememwm.train_negatives.enabled=true +- ++algorithm.dememwm.train_negatives.prob=0.10 +- ++algorithm.dememwm.train_negatives.start_step=20000 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=none +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=dit_patch +- ++algorithm.dememwm.revisit.compress.downsample_ratio=1 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=10000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=4 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=10000 +- experiment.training.max_steps=750000 diff --git a/outputs/2026-06-12/00-48-24/main.log b/outputs/2026-06-12/00-48-24/main.log new file mode 100644 index 0000000000000000000000000000000000000000..f1404cc929fdb52f266332807e2f3538ee118826 --- /dev/null +++ b/outputs/2026-06-12/00-48-24/main.log @@ -0,0 +1,5 @@ +[2026-06-12 00:49:22,766][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-12 00:49:43,664][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:49,278][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 00:48:19,029][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-15 00:48:19,413][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/00-48-25/.hydra/config.yaml b/outputs/2026-06-12/00-48-25/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..aa51822707170f2c75fc5292f065c9648969e23c --- /dev/null +++ b/outputs/2026-06-12/00-48-25/.hydra/config.yaml @@ -0,0 +1,262 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 750000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 10000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 10000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + training_sampling: + revisit_biased_prob: 0.3 + min_gap: 32 + candidate_starts: 8 + train_negatives: + enabled: true + prob: 0.1 + start_step: 20000 + branches: + - wrong_pose + - time_shuffle + - source_matched_random + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: none + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 20000 + ramp_steps: 20000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + snr_shifted_memory: + enabled: true + mode: step_offset + offset_steps: 1 + apply_anchor: true + apply_dynamic: false + apply_revisit: true + noise_per_target: true + mark_as_generated: false + revisit_metric: + enabled: true + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + injection: + memory_hidden_size: 512 + pose_aware: true + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_gate_none_snr_shift_lr8e5_8a100_bs4_global32_750k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_lr8e5_8a100_bs4_global32_750k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/00-48-25/.hydra/hydra.yaml b/outputs/2026-06-12/00-48-25/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..97133726d4390a9b2a93d3087c34421de30a1b52 --- /dev/null +++ b/outputs/2026-06-12/00-48-25/.hydra/hydra.yaml @@ -0,0 +1,249 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_gate_none_snr_shift_lr8e5_8a100_bs4_global32_750k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_lr8e5_8a100_bs4_global32_750k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.injection.pose_aware=true + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=20000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.snr_shifted_memory.enabled=true + - ++algorithm.dememwm.snr_shifted_memory.mode=step_offset + - ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 + - ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true + - ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false + - ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true + - ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true + - ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false + - ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 + - ++algorithm.dememwm.training_sampling.min_gap=32 + - ++algorithm.dememwm.train_negatives.enabled=true + - ++algorithm.dememwm.train_negatives.prob=0.10 + - ++algorithm.dememwm.train_negatives.start_step=20000 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=none + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=10000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=4 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=10000 + - experiment.training.max_steps=750000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=10000,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=20000,++algorithm.dememwm.generated_history_proxy.start_step=20000,++algorithm.dememwm.injection.pose_aware=true,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=none,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.snr_shifted_memory.apply_anchor=true,++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false,++algorithm.dememwm.snr_shifted_memory.apply_revisit=true,++algorithm.dememwm.snr_shifted_memory.enabled=true,++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false,++algorithm.dememwm.snr_shifted_memory.mode=step_offset,++algorithm.dememwm.snr_shifted_memory.noise_per_target=true,++algorithm.dememwm.snr_shifted_memory.offset_steps=1,++algorithm.dememwm.train_negatives.enabled=true,++algorithm.dememwm.train_negatives.prob=0.10,++algorithm.dememwm.train_negatives.start_step=20000,++algorithm.dememwm.training_sampling.min_gap=32,++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_gate_none_snr_shift_lr8e5_8a100_bs4_global32_750k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_lr8e5_8a100_bs4_global32_750k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=4,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=750000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=10000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/00-48-25 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/00-48-25/.hydra/overrides.yaml b/outputs/2026-06-12/00-48-25/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c4014a7d99ff6a70b2b8f002c479c5b7f21978a1 --- /dev/null +++ b/outputs/2026-06-12/00-48-25/.hydra/overrides.yaml @@ -0,0 +1,91 @@ +- +name=train_dememwm_sep_full_seq_gate_none_snr_shift_lr8e5_8a100_bs4_global32_750k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_lr8e5_8a100_bs4_global32_750k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.injection.pose_aware=true +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=20000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.snr_shifted_memory.enabled=true +- ++algorithm.dememwm.snr_shifted_memory.mode=step_offset +- ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 +- ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true +- ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false +- ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true +- ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true +- ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false +- ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 +- ++algorithm.dememwm.training_sampling.min_gap=32 +- ++algorithm.dememwm.train_negatives.enabled=true +- ++algorithm.dememwm.train_negatives.prob=0.10 +- ++algorithm.dememwm.train_negatives.start_step=20000 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=none +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=10000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=4 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=10000 +- experiment.training.max_steps=750000 diff --git a/outputs/2026-06-12/00-48-25/main.log b/outputs/2026-06-12/00-48-25/main.log new file mode 100644 index 0000000000000000000000000000000000000000..7c191da84eefd6a7239f20b60ea5c921ac767d6a --- /dev/null +++ b/outputs/2026-06-12/00-48-25/main.log @@ -0,0 +1,106 @@ +[2026-06-12 00:49:16,635][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-12 00:49:16,821][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-12 00:49:16,821][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-12 00:49:16,821][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-12 00:49:16,821][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-12 00:49:16,821][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-12 00:49:19,282][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 00:49:20,227][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 00:49:22,749][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-12 00:49:22,762][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-12 00:49:22,763][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-12 00:49:22,765][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-12 00:49:22,770][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-12 00:49:22,778][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-12 00:49:22,801][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-12 00:49:28,124][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-12 00:49:43,664][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:43,664][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:43,664][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:43,664][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:43,664][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:43,664][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:43,664][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:47,277][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:47,277][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:47,277][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:47,277][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:47,277][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:47,278][lightning.pytorch.trainer.connectors.signal_c 2.5 M +7 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +836 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,272.513 Total estimated model params size (MB) +[2026-06-12 00:49:49,278][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:49,279][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:49,279][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:49,279][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:49,279][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:49,279][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:49,279][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 03:44:30,912][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_pairs', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,922][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_revisit_frames', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,922][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,922][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/non_revisit_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,922][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,922][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,923][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_minus_non_revisit_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,923][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,923][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,923][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_psnr_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,924][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,924][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,924][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,924][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_psnr_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,924][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,925][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,925][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,925][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_psnr_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:44:30,925][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-15 00:48:19,029][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-15 00:48:19,028][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-15 00:48:19,029][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-15 00:48:19,029][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-15 00:48:19,029][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-15 00:48:19,029][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-15 00:48:19,029][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-15 00:48:19,115][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:48:19,119][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:48:19,400][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:48:19,406][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:48:20,380][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:48:20,386][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:48:20,391][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:48:19,035][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-15 00:48:19,030][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-15 00:48:19,035][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:48:19,035][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-15 00:48:19,035][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:48:19,035][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:48:19,030][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-15 00:48:19,121][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/00-48-26/.hydra/config.yaml b/outputs/2026-06-12/00-48-26/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..aa51822707170f2c75fc5292f065c9648969e23c --- /dev/null +++ b/outputs/2026-06-12/00-48-26/.hydra/config.yaml @@ -0,0 +1,262 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 750000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 10000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 10000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + training_sampling: + revisit_biased_prob: 0.3 + min_gap: 32 + candidate_starts: 8 + train_negatives: + enabled: true + prob: 0.1 + start_step: 20000 + branches: + - wrong_pose + - time_shuffle + - source_matched_random + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: none + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 20000 + ramp_steps: 20000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + snr_shifted_memory: + enabled: true + mode: step_offset + offset_steps: 1 + apply_anchor: true + apply_dynamic: false + apply_revisit: true + noise_per_target: true + mark_as_generated: false + revisit_metric: + enabled: true + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + injection: + memory_hidden_size: 512 + pose_aware: true + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_gate_none_snr_shift_lr8e5_8a100_bs4_global32_750k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_lr8e5_8a100_bs4_global32_750k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/00-48-26/.hydra/hydra.yaml b/outputs/2026-06-12/00-48-26/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..39a61532c360651afa124af25f4f6a5f614901ef --- /dev/null +++ b/outputs/2026-06-12/00-48-26/.hydra/hydra.yaml @@ -0,0 +1,249 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_gate_none_snr_shift_lr8e5_8a100_bs4_global32_750k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_lr8e5_8a100_bs4_global32_750k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.injection.pose_aware=true + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=20000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.snr_shifted_memory.enabled=true + - ++algorithm.dememwm.snr_shifted_memory.mode=step_offset + - ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 + - ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true + - ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false + - ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true + - ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true + - ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false + - ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 + - ++algorithm.dememwm.training_sampling.min_gap=32 + - ++algorithm.dememwm.train_negatives.enabled=true + - ++algorithm.dememwm.train_negatives.prob=0.10 + - ++algorithm.dememwm.train_negatives.start_step=20000 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=none + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=10000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=4 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=10000 + - experiment.training.max_steps=750000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=10000,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=20000,++algorithm.dememwm.generated_history_proxy.start_step=20000,++algorithm.dememwm.injection.pose_aware=true,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=none,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.snr_shifted_memory.apply_anchor=true,++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false,++algorithm.dememwm.snr_shifted_memory.apply_revisit=true,++algorithm.dememwm.snr_shifted_memory.enabled=true,++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false,++algorithm.dememwm.snr_shifted_memory.mode=step_offset,++algorithm.dememwm.snr_shifted_memory.noise_per_target=true,++algorithm.dememwm.snr_shifted_memory.offset_steps=1,++algorithm.dememwm.train_negatives.enabled=true,++algorithm.dememwm.train_negatives.prob=0.10,++algorithm.dememwm.train_negatives.start_step=20000,++algorithm.dememwm.training_sampling.min_gap=32,++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_gate_none_snr_shift_lr8e5_8a100_bs4_global32_750k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_lr8e5_8a100_bs4_global32_750k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=4,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=750000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=10000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/00-48-26 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/00-48-26/.hydra/overrides.yaml b/outputs/2026-06-12/00-48-26/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c4014a7d99ff6a70b2b8f002c479c5b7f21978a1 --- /dev/null +++ b/outputs/2026-06-12/00-48-26/.hydra/overrides.yaml @@ -0,0 +1,91 @@ +- +name=train_dememwm_sep_full_seq_gate_none_snr_shift_lr8e5_8a100_bs4_global32_750k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_lr8e5_8a100_bs4_global32_750k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.injection.pose_aware=true +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=20000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.snr_shifted_memory.enabled=true +- ++algorithm.dememwm.snr_shifted_memory.mode=step_offset +- ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 +- ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true +- ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false +- ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true +- ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true +- ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false +- ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 +- ++algorithm.dememwm.training_sampling.min_gap=32 +- ++algorithm.dememwm.train_negatives.enabled=true +- ++algorithm.dememwm.train_negatives.prob=0.10 +- ++algorithm.dememwm.train_negatives.start_step=20000 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=none +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=10000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=4 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=10000 +- experiment.training.max_steps=750000 diff --git a/outputs/2026-06-12/00-48-26/main.log b/outputs/2026-06-12/00-48-26/main.log new file mode 100644 index 0000000000000000000000000000000000000000..79a76c8a5bac7bf8d7b6264891fcc5bf389b0522 --- /dev/null +++ b/outputs/2026-06-12/00-48-26/main.log @@ -0,0 +1,79 @@ +[2026-06-12 00:49:15,159][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-12 00:49:15,353][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-12 00:49:15,353][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-12 00:49:15,353][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-12 00:49:15,353][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-12 00:49:15,353][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-12 00:49:18,355][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 00:49:19,243][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 00:49:21,767][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-12 00:49:21,775][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-12 00:49:27,831][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-12 00:49:43,209][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:43,209][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:49:44,184][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 834 M +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 887 K +2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 296 K +3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 296 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | dememwm_memory_pose_embed | Sequential | 266 K +6 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +7 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +836 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,273.565 Total estimated model params size (MB) +[2026-06-12 00:49:47,277][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:49:47,277][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 03:48:30,553][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_pairs', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,361][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_revisit_frames', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,362][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,362][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/non_revisit_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,362][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,363][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,363][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_minus_non_revisit_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,363][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,363][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,363][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_psnr_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,364][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,364][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,364][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,364][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_psnr_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,365][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,365][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,365][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,365][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_psnr_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 03:48:31,366][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-15 00:48:19,030][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-15 00:48:19,035][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:48:19,029][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-15 00:48:19,035][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-15 00:48:19,036][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:48:19,799][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/00-49-54/.hydra/config.yaml b/outputs/2026-06-12/00-49-54/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..55d20637a0061a8b4a79eab4fbd188633caa8bbb --- /dev/null +++ b/outputs/2026-06-12/00-49-54/.hydra/config.yaml @@ -0,0 +1,262 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 750000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 10000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 10000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + training_sampling: + revisit_biased_prob: 0.3 + min_gap: 32 + candidate_starts: 8 + train_negatives: + enabled: true + prob: 0.1 + start_step: 20000 + branches: + - wrong_pose + - time_shuffle + - source_matched_random + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: none + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 20000 + ramp_steps: 20000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + snr_shifted_memory: + enabled: true + mode: step_offset + offset_steps: 1 + apply_anchor: true + apply_dynamic: false + apply_revisit: true + noise_per_target: true + mark_as_generated: false + revisit_metric: + enabled: true + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + injection: + memory_hidden_size: 512 + pose_aware: true + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/00-49-54/.hydra/hydra.yaml b/outputs/2026-06-12/00-49-54/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..34eba634fa847e5c8e40447ac5fbd772d1f34eda --- /dev/null +++ b/outputs/2026-06-12/00-49-54/.hydra/hydra.yaml @@ -0,0 +1,249 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.injection.pose_aware=true + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=20000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.snr_shifted_memory.enabled=true + - ++algorithm.dememwm.snr_shifted_memory.mode=step_offset + - ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 + - ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true + - ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false + - ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true + - ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true + - ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false + - ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 + - ++algorithm.dememwm.training_sampling.min_gap=32 + - ++algorithm.dememwm.train_negatives.enabled=true + - ++algorithm.dememwm.train_negatives.prob=0.10 + - ++algorithm.dememwm.train_negatives.start_step=20000 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=none + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=10000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=4 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=10000 + - experiment.training.max_steps=750000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=10000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=20000,++algorithm.dememwm.generated_history_proxy.start_step=20000,++algorithm.dememwm.injection.pose_aware=true,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=none,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.snr_shifted_memory.apply_anchor=true,++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false,++algorithm.dememwm.snr_shifted_memory.apply_revisit=true,++algorithm.dememwm.snr_shifted_memory.enabled=true,++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false,++algorithm.dememwm.snr_shifted_memory.mode=step_offset,++algorithm.dememwm.snr_shifted_memory.noise_per_target=true,++algorithm.dememwm.snr_shifted_memory.offset_steps=1,++algorithm.dememwm.train_negatives.enabled=true,++algorithm.dememwm.train_negatives.prob=0.10,++algorithm.dememwm.train_negatives.start_step=20000,++algorithm.dememwm.training_sampling.min_gap=32,++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=4,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=750000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=10000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/00-49-54 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/00-49-54/.hydra/overrides.yaml b/outputs/2026-06-12/00-49-54/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..80caf7ad361473e24bf1e74f22bf1d873b276fbe --- /dev/null +++ b/outputs/2026-06-12/00-49-54/.hydra/overrides.yaml @@ -0,0 +1,91 @@ +- +name=train_dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.injection.pose_aware=true +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=20000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.snr_shifted_memory.enabled=true +- ++algorithm.dememwm.snr_shifted_memory.mode=step_offset +- ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 +- ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true +- ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false +- ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true +- ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true +- ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false +- ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 +- ++algorithm.dememwm.training_sampling.min_gap=32 +- ++algorithm.dememwm.train_negatives.enabled=true +- ++algorithm.dememwm.train_negatives.prob=0.10 +- ++algorithm.dememwm.train_negatives.start_step=20000 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=none +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=10000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=4 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=10000 +- experiment.training.max_steps=750000 diff --git a/outputs/2026-06-12/00-49-54/main.log b/outputs/2026-06-12/00-49-54/main.log new file mode 100644 index 0000000000000000000000000000000000000000..16aee1c7d8486b71cb2020f5bbdc293deec5e70d --- /dev/null +++ b/outputs/2026-06-12/00-49-54/main.log @@ -0,0 +1,92 @@ +[2026-06-12 00:50:39,167][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-12 00:50:39,538][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-12 00:50:39,538][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-12 00:50:39,538][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-12 00:50:39,538][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-12 00:50:39,539][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-12 00:50:43,264][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 00:50:44,227][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 00:50:46,806][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-12 00:50:46,806][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-12 00:50:46,806][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-12 00:50:46,806][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-12 00:50:46,806][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-12 00:50:52,936][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-12 00:51:06,875][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:51:06,875][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:51:06,875][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:51:06,875][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:51:06,875][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:51:08,475][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 834 M +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 887 K +2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 296 K +3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 296 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | dememwm_memory_pose_embed | Sequential | 266 K +6 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +7 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +836 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,273.565 Total estimated model params size (MB) +[2026-06-12 00:51:09,477][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:51:09,477][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:51:09,477][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:51:09,477][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:51:09,478][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 07:47:02,486][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_pairs', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,381][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_revisit_frames', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,381][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,382][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/non_revisit_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,382][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,382][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,383][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_minus_non_revisit_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,383][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,383][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,383][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_psnr_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,384][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,384][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,384][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,384][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_psnr_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,385][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,385][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,385][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,385][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_psnr_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-12 07:47:03,385][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-15 00:49:48,753][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-15 00:49:48,754][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-15 00:49:48,755][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:49:48,755][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:49:48,754][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-15 00:49:48,755][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:49:48,755][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-15 00:49:48,754][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-15 00:49:48,755][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:49:48,756][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/00-49-55/.hydra/config.yaml b/outputs/2026-06-12/00-49-55/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..55d20637a0061a8b4a79eab4fbd188633caa8bbb --- /dev/null +++ b/outputs/2026-06-12/00-49-55/.hydra/config.yaml @@ -0,0 +1,262 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 750000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 10000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 10000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + training_sampling: + revisit_biased_prob: 0.3 + min_gap: 32 + candidate_starts: 8 + train_negatives: + enabled: true + prob: 0.1 + start_step: 20000 + branches: + - wrong_pose + - time_shuffle + - source_matched_random + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: none + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 20000 + ramp_steps: 20000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + snr_shifted_memory: + enabled: true + mode: step_offset + offset_steps: 1 + apply_anchor: true + apply_dynamic: false + apply_revisit: true + noise_per_target: true + mark_as_generated: false + revisit_metric: + enabled: true + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + injection: + memory_hidden_size: 512 + pose_aware: true + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/00-49-55/.hydra/hydra.yaml b/outputs/2026-06-12/00-49-55/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ae1347d6c54c8a55dedbc4898f1a7951a3d4916d --- /dev/null +++ b/outputs/2026-06-12/00-49-55/.hydra/hydra.yaml @@ -0,0 +1,249 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.injection.pose_aware=true + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=20000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.snr_shifted_memory.enabled=true + - ++algorithm.dememwm.snr_shifted_memory.mode=step_offset + - ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 + - ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true + - ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false + - ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true + - ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true + - ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false + - ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 + - ++algorithm.dememwm.training_sampling.min_gap=32 + - ++algorithm.dememwm.train_negatives.enabled=true + - ++algorithm.dememwm.train_negatives.prob=0.10 + - ++algorithm.dememwm.train_negatives.start_step=20000 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=none + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=10000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=4 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=10000 + - experiment.training.max_steps=750000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=10000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=20000,++algorithm.dememwm.generated_history_proxy.start_step=20000,++algorithm.dememwm.injection.pose_aware=true,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=none,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.snr_shifted_memory.apply_anchor=true,++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false,++algorithm.dememwm.snr_shifted_memory.apply_revisit=true,++algorithm.dememwm.snr_shifted_memory.enabled=true,++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false,++algorithm.dememwm.snr_shifted_memory.mode=step_offset,++algorithm.dememwm.snr_shifted_memory.noise_per_target=true,++algorithm.dememwm.snr_shifted_memory.offset_steps=1,++algorithm.dememwm.train_negatives.enabled=true,++algorithm.dememwm.train_negatives.prob=0.10,++algorithm.dememwm.train_negatives.start_step=20000,++algorithm.dememwm.training_sampling.min_gap=32,++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=4,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=750000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=10000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/00-49-55 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/00-49-55/.hydra/overrides.yaml b/outputs/2026-06-12/00-49-55/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..80caf7ad361473e24bf1e74f22bf1d873b276fbe --- /dev/null +++ b/outputs/2026-06-12/00-49-55/.hydra/overrides.yaml @@ -0,0 +1,91 @@ +- +name=train_dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.injection.pose_aware=true +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=20000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.snr_shifted_memory.enabled=true +- ++algorithm.dememwm.snr_shifted_memory.mode=step_offset +- ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 +- ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true +- ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false +- ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true +- ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true +- ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false +- ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 +- ++algorithm.dememwm.training_sampling.min_gap=32 +- ++algorithm.dememwm.train_negatives.enabled=true +- ++algorithm.dememwm.train_negatives.prob=0.10 +- ++algorithm.dememwm.train_negatives.start_step=20000 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=none +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=10000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=4 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=10000 +- experiment.training.max_steps=750000 diff --git a/outputs/2026-06-12/00-49-55/main.log b/outputs/2026-06-12/00-49-55/main.log new file mode 100644 index 0000000000000000000000000000000000000000..d1382db40f796e0c9b60ef0641fed449454c40f4 --- /dev/null +++ b/outputs/2026-06-12/00-49-55/main.log @@ -0,0 +1,15 @@ +[2026-06-12 00:50:46,806][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-12 00:50:46,806][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-12 00:50:46,807][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-12 00:51:06,875][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:51:06,875][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:51:06,875][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 00:51:09,477][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:51:09,477][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 00:51:09,478][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 00:49:48,753][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-15 00:49:48,755][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:49:48,755][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-15 00:49:48,755][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 00:49:48,753][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-15 00:49:48,757][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/09-13-36/.hydra/config.yaml b/outputs/2026-06-12/09-13-36/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b47e3332559c53ed999c275a717b07d1982311f8 --- /dev/null +++ b/outputs/2026-06-12/09-13-36/.hydra/config.yaml @@ -0,0 +1,231 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: + - 3 + - 7 + - 11 + - 15 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/09-13-36/.hydra/hydra.yaml b/outputs/2026-06-12/09-13-36/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b45a7c94f520e55c7be18b19ae5865a470b99e34 --- /dev/null +++ b/outputs/2026-06-12/09-13-36/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - algorithm.memory_cross_attn_layers=[3,7,11,15] + - ++algorithm.dememwm.injection.memory_hidden_size=512 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.injection.memory_hidden_size=512,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.memory_cross_attn_layers=[3,7,11,15],algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/09-13-36 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/09-13-36/.hydra/overrides.yaml b/outputs/2026-06-12/09-13-36/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8ee3288e706e04827ca476ead0592fea9c288ffa --- /dev/null +++ b/outputs/2026-06-12/09-13-36/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- algorithm.memory_cross_attn_layers=[3,7,11,15] +- ++algorithm.dememwm.injection.memory_hidden_size=512 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-12/09-13-36/main.log b/outputs/2026-06-12/09-13-36/main.log new file mode 100644 index 0000000000000000000000000000000000000000..4a33adeba034db40570e99dc4501678f76cc3f5f --- /dev/null +++ b/outputs/2026-06-12/09-13-36/main.log @@ -0,0 +1,77 @@ +[2026-06-12 09:14:21,877][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-12 09:14:21,992][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-12 09:14:21,993][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-12 09:14:21,993][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-12 09:14:21,993][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-12 09:14:21,993][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-12 09:14:24,614][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 09:14:25,556][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 09:14:28,274][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-12 09:14:28,275][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-12 09:14:28,275][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-12 09:14:28,275][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-12 09:14:28,275][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-12 09:14:28,275][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-12 09:14:28,275][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-12 09:14:34,383][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-12 09:14:48,621][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 09:14:48,622][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 09:14:48,622][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 09:14:48,622][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 09:14:48,622][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 09:14:48,622][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 09:14:48,622][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 09:14:50,305][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 681 M +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 887 K +2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 296 K +3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 296 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +6 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +682 M Trainable params +231 M Non-trainable params +914 M Total params +3,658.543 Total estimated model params size (MB) +[2026-06-12 09:14:51,296][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 09:14:51,296][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 09:14:51,296][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 09:14:51,296][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 09:14:51,296][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 09:14:51,296][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 09:14:51,296][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 09:13:48,033][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-15 09:13:48,033][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-15 09:13:48,034][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 09:13:48,034][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 09:13:48,033][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-15 09:13:48,033][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-15 09:13:48,034][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 09:13:48,034][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 09:13:48,033][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-15 09:13:48,034][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 09:13:48,033][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-15 09:13:48,035][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 09:13:48,036][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-15 09:13:48,036][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 09:13:48,040][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-15 09:13:48,040][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 09:13:48,040][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-15 09:13:48,041][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 09:13:48,041][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-15 09:13:48,041][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 09:13:48,041][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-15 09:13:48,041][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 09:13:48,041][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-15 09:13:48,041][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 09:13:48,042][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-15 09:13:48,042][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 09:13:48,043][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-15 09:13:48,043][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/09-13-37/.hydra/config.yaml b/outputs/2026-06-12/09-13-37/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b47e3332559c53ed999c275a717b07d1982311f8 --- /dev/null +++ b/outputs/2026-06-12/09-13-37/.hydra/config.yaml @@ -0,0 +1,231 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: + - 3 + - 7 + - 11 + - 15 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/09-13-37/.hydra/hydra.yaml b/outputs/2026-06-12/09-13-37/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5832f21c92f34fd5a59de0c06a8710d85595e17a --- /dev/null +++ b/outputs/2026-06-12/09-13-37/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - algorithm.memory_cross_attn_layers=[3,7,11,15] + - ++algorithm.dememwm.injection.memory_hidden_size=512 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.injection.memory_hidden_size=512,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.memory_cross_attn_layers=[3,7,11,15],algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/09-13-37 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/09-13-37/.hydra/overrides.yaml b/outputs/2026-06-12/09-13-37/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8ee3288e706e04827ca476ead0592fea9c288ffa --- /dev/null +++ b/outputs/2026-06-12/09-13-37/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- algorithm.memory_cross_attn_layers=[3,7,11,15] +- ++algorithm.dememwm.injection.memory_hidden_size=512 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-12/09-13-37/main.log b/outputs/2026-06-12/09-13-37/main.log new file mode 100644 index 0000000000000000000000000000000000000000..ef2f4e3653893acee7d9a33fcc4dec4894caaf7f --- /dev/null +++ b/outputs/2026-06-12/09-13-37/main.log @@ -0,0 +1,7 @@ +[2026-06-12 09:14:28,275][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-12 09:14:48,622][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 09:14:51,296][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 09:13:48,033][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-15 09:13:48,034][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 09:13:48,040][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-15 09:13:48,040][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/19-27-06/.hydra/config.yaml b/outputs/2026-06-12/19-27-06/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0407b139e7d545e5e741be674e569dea342b0ccc --- /dev/null +++ b/outputs/2026-06-12/19-27-06/.hydra/config.yaml @@ -0,0 +1,227 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 1024 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/19-27-06/.hydra/hydra.yaml b/outputs/2026-06-12/19-27-06/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4bc2ba09df8e9cd3062832277abcf52f3b78570f --- /dev/null +++ b/outputs/2026-06-12/19-27-06/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - algorithm.memory_cross_attn_layers=null + - ++algorithm.dememwm.injection.memory_hidden_size=1024 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.injection.memory_hidden_size=1024,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.memory_cross_attn_layers=null,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/19-27-06 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/19-27-06/.hydra/overrides.yaml b/outputs/2026-06-12/19-27-06/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4cb6f8a5aeb033b8dea3cbf193d9bd78ce5d0b0b --- /dev/null +++ b/outputs/2026-06-12/19-27-06/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- algorithm.memory_cross_attn_layers=null +- ++algorithm.dememwm.injection.memory_hidden_size=1024 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-12/19-27-06/main.log b/outputs/2026-06-12/19-27-06/main.log new file mode 100644 index 0000000000000000000000000000000000000000..102dd2f9b4cd3e131da305bb037d9b691f869baa --- /dev/null +++ b/outputs/2026-06-12/19-27-06/main.log @@ -0,0 +1,53 @@ +[2026-06-12 19:27:53,749][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-12 19:27:54,699][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-12 19:27:54,699][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-12 19:27:54,699][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-12 19:27:54,699][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-12 19:27:54,700][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-12 19:27:57,364][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 19:27:58,336][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 19:28:01,170][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-12 19:28:01,171][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-12 19:28:01,184][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-12 19:28:01,186][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-12 19:28:01,214][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-12 19:28:06,789][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-12 19:28:21,673][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 19:28:21,673][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 19:28:21,673][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 19:28:21,673][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 19:28:21,673][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 19:28:23,087][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 1.8 M +2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 591 K +3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 591 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +6 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +1.0 B Trainable params +231 M Non-trainable params +1.2 B Total params +4,983.118 Total estimated model params size (MB) +[2026-06-12 19:28:24,095][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 19:28:24,096][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 19:28:24,095][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 19:28:24,095][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 19:28:24,095][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 19:27:23,331][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-15 19:27:23,332][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 19:27:23,332][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-15 19:27:23,333][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 19:27:23,332][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-15 19:27:23,333][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 19:27:23,332][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-15 19:27:23,332][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-15 19:27:23,333][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 19:27:23,333][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/19-27-07/.hydra/config.yaml b/outputs/2026-06-12/19-27-07/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0407b139e7d545e5e741be674e569dea342b0ccc --- /dev/null +++ b/outputs/2026-06-12/19-27-07/.hydra/config.yaml @@ -0,0 +1,227 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 1024 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/19-27-07/.hydra/hydra.yaml b/outputs/2026-06-12/19-27-07/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4d5a789479e8a19a80c6168dcc93961975417d43 --- /dev/null +++ b/outputs/2026-06-12/19-27-07/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - algorithm.memory_cross_attn_layers=null + - ++algorithm.dememwm.injection.memory_hidden_size=1024 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.injection.memory_hidden_size=1024,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.memory_cross_attn_layers=null,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/19-27-07 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/19-27-07/.hydra/overrides.yaml b/outputs/2026-06-12/19-27-07/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4cb6f8a5aeb033b8dea3cbf193d9bd78ce5d0b0b --- /dev/null +++ b/outputs/2026-06-12/19-27-07/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- algorithm.memory_cross_attn_layers=null +- ++algorithm.dememwm.injection.memory_hidden_size=1024 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-12/19-27-07/main.log b/outputs/2026-06-12/19-27-07/main.log new file mode 100644 index 0000000000000000000000000000000000000000..f22b75e24b37d29186b119ddee629c41d93efcda --- /dev/null +++ b/outputs/2026-06-12/19-27-07/main.log @@ -0,0 +1,10 @@ +[2026-06-12 19:28:01,175][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-12 19:28:01,185][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-12 19:28:21,673][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 19:28:21,673][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 19:28:24,095][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 19:28:24,095][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 19:27:23,332][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-15 19:27:23,333][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 19:27:23,332][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-15 19:27:23,333][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/19-27-08/.hydra/config.yaml b/outputs/2026-06-12/19-27-08/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0407b139e7d545e5e741be674e569dea342b0ccc --- /dev/null +++ b/outputs/2026-06-12/19-27-08/.hydra/config.yaml @@ -0,0 +1,227 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 1024 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/19-27-08/.hydra/hydra.yaml b/outputs/2026-06-12/19-27-08/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8926ceccd522a17c9d099df6b63e4ca5958005c7 --- /dev/null +++ b/outputs/2026-06-12/19-27-08/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - algorithm.memory_cross_attn_layers=null + - ++algorithm.dememwm.injection.memory_hidden_size=1024 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.injection.memory_hidden_size=1024,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.memory_cross_attn_layers=null,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/19-27-08 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/19-27-08/.hydra/overrides.yaml b/outputs/2026-06-12/19-27-08/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4cb6f8a5aeb033b8dea3cbf193d9bd78ce5d0b0b --- /dev/null +++ b/outputs/2026-06-12/19-27-08/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_alllayers1024_valid500_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- algorithm.memory_cross_attn_layers=null +- ++algorithm.dememwm.injection.memory_hidden_size=1024 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-12/19-27-08/main.log b/outputs/2026-06-12/19-27-08/main.log new file mode 100644 index 0000000000000000000000000000000000000000..77e2ea4ebf7cf63e135bb8299f31481c202af4fd --- /dev/null +++ b/outputs/2026-06-12/19-27-08/main.log @@ -0,0 +1,5 @@ +[2026-06-12 19:28:01,187][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-12 19:28:21,673][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 19:28:24,095][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 19:27:23,332][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-15 19:27:23,333][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/20-29-00/.hydra/config.yaml b/outputs/2026-06-12/20-29-00/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c604bc2394aaab5ec435addbaf49b8ad5d6093c2 --- /dev/null +++ b/outputs/2026-06-12/20-29-00/.hydra/config.yaml @@ -0,0 +1,230 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 1100 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: learned + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_base_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_base_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/20-29-00/.hydra/hydra.yaml b/outputs/2026-06-12/20-29-00/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..900382e1f4ee5605101ff311f67612e03f840701 --- /dev/null +++ b/outputs/2026-06-12/20-29-00/.hydra/hydra.yaml @@ -0,0 +1,233 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_base_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_base_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=1100 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=learned + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=learned,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=1100,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_base_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_base_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/20-29-00 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/20-29-00/.hydra/overrides.yaml b/outputs/2026-06-12/20-29-00/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c5f3775fa2a65fb67c1ac07a2800951036ac581a --- /dev/null +++ b/outputs/2026-06-12/20-29-00/.hydra/overrides.yaml @@ -0,0 +1,75 @@ +- +name=train_dememwm_sep_full_seq_base_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_base_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=1100 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=learned +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-12/20-29-00/main.log b/outputs/2026-06-12/20-29-00/main.log new file mode 100644 index 0000000000000000000000000000000000000000..031f1442ccf02665b5705a8c830564b41f6f6430 --- /dev/null +++ b/outputs/2026-06-12/20-29-00/main.log @@ -0,0 +1,25 @@ +[2026-06-12 20:29:51,095][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-12 20:29:51,096][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-12 20:29:51,096][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-12 20:29:51,097][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-12 20:29:51,102][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-12 20:30:12,541][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:30:12,541][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:30:12,541][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:30:12,541][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:30:12,541][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:30:16,982][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 20:30:16,982][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 20:30:16,982][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 20:30:16,982][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 20:30:16,982][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 20:28:55,038][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-15 20:28:55,038][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-15 20:28:55,039][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-15 20:28:55,039][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-15 20:28:55,039][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-15 20:28:55,269][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:28:55,275][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:28:55,282][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:28:56,289][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:28:57,242][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/20-29-02/.hydra/config.yaml b/outputs/2026-06-12/20-29-02/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c604bc2394aaab5ec435addbaf49b8ad5d6093c2 --- /dev/null +++ b/outputs/2026-06-12/20-29-02/.hydra/config.yaml @@ -0,0 +1,230 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 1100 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: learned + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_base_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_base_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/20-29-02/.hydra/hydra.yaml b/outputs/2026-06-12/20-29-02/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..6a54788a2de76a737bb0802b04a31c5008f40c6e --- /dev/null +++ b/outputs/2026-06-12/20-29-02/.hydra/hydra.yaml @@ -0,0 +1,233 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_base_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_base_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=1100 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=learned + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=learned,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=1100,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_base_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_base_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/20-29-02 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/20-29-02/.hydra/overrides.yaml b/outputs/2026-06-12/20-29-02/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c5f3775fa2a65fb67c1ac07a2800951036ac581a --- /dev/null +++ b/outputs/2026-06-12/20-29-02/.hydra/overrides.yaml @@ -0,0 +1,75 @@ +- +name=train_dememwm_sep_full_seq_base_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_base_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=1100 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=learned +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-12/20-29-02/main.log b/outputs/2026-06-12/20-29-02/main.log new file mode 100644 index 0000000000000000000000000000000000000000..f73b1074d1dd3156b62ff3236188165402710675 --- /dev/null +++ b/outputs/2026-06-12/20-29-02/main.log @@ -0,0 +1,33 @@ +[2026-06-12 20:29:43,958][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-12 20:29:44,123][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-12 20:29:44,123][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-12 20:29:44,123][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-12 20:29:44,124][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-12 20:29:44,124][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-12 20:29:47,170][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 20:29:48,143][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 20:29:51,096][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-12 20:29:57,625][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-12 20:30:12,541][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:30:14,990][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 834 M +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 887 K +2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 296 K +3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 296 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +6 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +836 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,271.812 Total estimated model params size (MB) +[2026-06-12 20:30:16,982][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 20:28:55,038][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-15 20:28:55,265][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/20-29-03/.hydra/config.yaml b/outputs/2026-06-12/20-29-03/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c604bc2394aaab5ec435addbaf49b8ad5d6093c2 --- /dev/null +++ b/outputs/2026-06-12/20-29-03/.hydra/config.yaml @@ -0,0 +1,230 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 1100 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: learned + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_base_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_base_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/20-29-03/.hydra/hydra.yaml b/outputs/2026-06-12/20-29-03/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4704299eef82eabbbc25f9b3280540a8119ae833 --- /dev/null +++ b/outputs/2026-06-12/20-29-03/.hydra/hydra.yaml @@ -0,0 +1,233 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_base_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_base_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=1100 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=learned + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=learned,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=1100,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_base_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_base_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/20-29-03 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/20-29-03/.hydra/overrides.yaml b/outputs/2026-06-12/20-29-03/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c5f3775fa2a65fb67c1ac07a2800951036ac581a --- /dev/null +++ b/outputs/2026-06-12/20-29-03/.hydra/overrides.yaml @@ -0,0 +1,75 @@ +- +name=train_dememwm_sep_full_seq_base_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_base_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=1100 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=learned +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-12/20-29-03/main.log b/outputs/2026-06-12/20-29-03/main.log new file mode 100644 index 0000000000000000000000000000000000000000..16122ddee07fc02f4393bbb4a59e4efee869a25b --- /dev/null +++ b/outputs/2026-06-12/20-29-03/main.log @@ -0,0 +1,10 @@ +[2026-06-12 20:29:51,097][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-12 20:29:51,097][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-12 20:30:12,541][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:30:12,541][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:30:16,982][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 20:30:16,982][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 20:28:55,039][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-15 20:28:55,039][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-15 20:28:55,144][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:28:56,330][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/20-31-35/.hydra/config.yaml b/outputs/2026-06-12/20-31-35/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..31a98d1e56829de50924f50c9e233702aa6798d3 --- /dev/null +++ b/outputs/2026-06-12/20-31-35/.hydra/config.yaml @@ -0,0 +1,237 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 1100 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: dit_patch + downsample_ratio: null + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: none + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: dit_patch + downsample_ratio: null + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + snr_shifted_memory: + enabled: false + mode: step_offset + offset_steps: 1 + apply_anchor: true + apply_dynamic: false + apply_revisit: true + noise_per_target: true + mark_as_generated: false + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_gate_none_nocomp_a2_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_nocomp_a2_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/20-31-35/.hydra/hydra.yaml b/outputs/2026-06-12/20-31-35/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..2051f3e54083ee28c5b99434be25319420e7da72 --- /dev/null +++ b/outputs/2026-06-12/20-31-35/.hydra/hydra.yaml @@ -0,0 +1,233 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_gate_none_nocomp_a2_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_nocomp_a2_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=1100 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=dit_patch + - ++algorithm.dememwm.anchor.compress.downsample_ratio=null + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=none + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=dit_patch + - ++algorithm.dememwm.revisit.compress.downsample_ratio=null + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1],++algorithm.dememwm.anchor.compress.downsample_ratio=null,++algorithm.dememwm.anchor.compress.tokenization=dit_patch,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=null,++algorithm.dememwm.revisit.compress.tokenization=dit_patch,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=none,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=1100,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_gate_none_nocomp_a2_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_nocomp_a2_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/20-31-35 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/20-31-35/.hydra/overrides.yaml b/outputs/2026-06-12/20-31-35/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0ce4a4f18540b037c78822c163519003195c0134 --- /dev/null +++ b/outputs/2026-06-12/20-31-35/.hydra/overrides.yaml @@ -0,0 +1,75 @@ +- +name=train_dememwm_sep_full_seq_gate_none_nocomp_a2_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_nocomp_a2_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=1100 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=dit_patch +- ++algorithm.dememwm.anchor.compress.downsample_ratio=null +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=none +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=dit_patch +- ++algorithm.dememwm.revisit.compress.downsample_ratio=null +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-12/20-31-35/main.log b/outputs/2026-06-12/20-31-35/main.log new file mode 100644 index 0000000000000000000000000000000000000000..13b4d2650e12a049bd0d7246f63d8d85b69781d8 --- /dev/null +++ b/outputs/2026-06-12/20-31-35/main.log @@ -0,0 +1,70 @@ +[2026-06-12 20:32:21,967][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-12 20:32:22,094][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-12 20:32:22,094][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-12 20:32:22,095][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-12 20:32:22,095][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-12 20:32:22,095][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-12 20:32:25,319][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 20:32:26,291][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 20:32:28,875][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-12 20:32:28,876][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-12 20:32:28,876][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-12 20:32:28,876][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-12 20:32:28,876][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-12 20:32:28,876][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-12 20:32:34,395][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-12 20:32:47,524][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:32:47,524][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:32:47,524][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:32:47,524][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:32:47,524][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:32:47,524][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:32:49,416][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 834 M +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 887 K +2 | dememwm_anchor_proj | DiTPatchMemoryProjector | 33.3 K +3 | dememwm_revisit_proj | DiTPatchMemoryProjector | 33.3 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +6 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +835 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,269.707 Total estimated model params size (MB) +[2026-06-12 20:32:53,400][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 20:32:53,400][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 20:32:53,400][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 20:32:53,400][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 20:32:53,400][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 20:32:53,400][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 20:31:25,040][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-15 20:31:25,040][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-15 20:31:25,041][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:31:25,041][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:31:25,041][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-15 20:31:25,040][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-15 20:31:25,041][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:31:25,041][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:31:25,041][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-15 20:31:25,041][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:31:25,040][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-15 20:31:25,043][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:31:25,045][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-15 20:31:25,045][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:31:25,045][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-15 20:31:25,045][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:31:25,046][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-15 20:31:25,046][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:31:25,046][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-15 20:31:25,046][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:31:25,046][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-15 20:31:25,046][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:31:25,046][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-15 20:31:25,046][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/20-31-36/.hydra/config.yaml b/outputs/2026-06-12/20-31-36/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..37657208acae3cddc363fbf170f94749a1917b00 --- /dev/null +++ b/outputs/2026-06-12/20-31-36/.hydra/config.yaml @@ -0,0 +1,239 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 1100 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: none + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + snr_shifted_memory: + enabled: true + mode: step_offset + offset_steps: 1 + apply_anchor: true + apply_dynamic: false + apply_revisit: true + noise_per_target: true + mark_as_generated: false + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_gate_none_snr_shift_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/20-31-36/.hydra/hydra.yaml b/outputs/2026-06-12/20-31-36/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c3b2bab34f7d924562b8438f41396e6b40e81455 --- /dev/null +++ b/outputs/2026-06-12/20-31-36/.hydra/hydra.yaml @@ -0,0 +1,241 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_gate_none_snr_shift_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=1100 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.snr_shifted_memory.enabled=true + - ++algorithm.dememwm.snr_shifted_memory.mode=step_offset + - ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 + - ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true + - ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false + - ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true + - ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true + - ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=none + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=none,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.snr_shifted_memory.apply_anchor=true,++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false,++algorithm.dememwm.snr_shifted_memory.apply_revisit=true,++algorithm.dememwm.snr_shifted_memory.enabled=true,++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false,++algorithm.dememwm.snr_shifted_memory.mode=step_offset,++algorithm.dememwm.snr_shifted_memory.noise_per_target=true,++algorithm.dememwm.snr_shifted_memory.offset_steps=1,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=1100,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_gate_none_snr_shift_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/20-31-36 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/20-31-36/.hydra/overrides.yaml b/outputs/2026-06-12/20-31-36/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..67a4320747586094d17d9591a2df050cfadcf719 --- /dev/null +++ b/outputs/2026-06-12/20-31-36/.hydra/overrides.yaml @@ -0,0 +1,83 @@ +- +name=train_dememwm_sep_full_seq_gate_none_snr_shift_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=1100 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.snr_shifted_memory.enabled=true +- ++algorithm.dememwm.snr_shifted_memory.mode=step_offset +- ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 +- ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true +- ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false +- ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true +- ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true +- ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=none +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-12/20-31-36/main.log b/outputs/2026-06-12/20-31-36/main.log new file mode 100644 index 0000000000000000000000000000000000000000..08be6205e59a7e2b64cf2e5d6bb0600981804376 Binary files /dev/null and b/outputs/2026-06-12/20-31-36/main.log differ diff --git a/outputs/2026-06-12/20-31-39/.hydra/config.yaml b/outputs/2026-06-12/20-31-39/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..969b5c001b3c5895df04636acb7922a0506e92c3 --- /dev/null +++ b/outputs/2026-06-12/20-31-39/.hydra/config.yaml @@ -0,0 +1,239 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 1100 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: none + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + snr_shifted_memory: + enabled: false + mode: step_offset + offset_steps: 1 + apply_anchor: true + apply_dynamic: false + apply_revisit: true + noise_per_target: true + mark_as_generated: false + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/20-31-39/.hydra/hydra.yaml b/outputs/2026-06-12/20-31-39/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9b4d2d34cdb8273078dbddc36de3b145113ffb8f --- /dev/null +++ b/outputs/2026-06-12/20-31-39/.hydra/hydra.yaml @@ -0,0 +1,233 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=1100 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=none + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=none,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=1100,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/20-31-39 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/20-31-39/.hydra/overrides.yaml b/outputs/2026-06-12/20-31-39/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3ea991f8fd7f385f8fc1231d04a8448120510272 --- /dev/null +++ b/outputs/2026-06-12/20-31-39/.hydra/overrides.yaml @@ -0,0 +1,75 @@ +- +name=train_dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=1100 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=none +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-12/20-31-39/main.log b/outputs/2026-06-12/20-31-39/main.log new file mode 100644 index 0000000000000000000000000000000000000000..9686260614187eacbb62aefe4a9a2a349fe316e1 --- /dev/null +++ b/outputs/2026-06-12/20-31-39/main.log @@ -0,0 +1,5 @@ +[2026-06-12 20:32:29,964][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-12 20:32:52,650][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:32:55,925][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 20:31:25,200][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-15 20:31:25,201][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/20-31-45/.hydra/config.yaml b/outputs/2026-06-12/20-31-45/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..969b5c001b3c5895df04636acb7922a0506e92c3 --- /dev/null +++ b/outputs/2026-06-12/20-31-45/.hydra/config.yaml @@ -0,0 +1,239 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 1100 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: none + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + snr_shifted_memory: + enabled: false + mode: step_offset + offset_steps: 1 + apply_anchor: true + apply_dynamic: false + apply_revisit: true + noise_per_target: true + mark_as_generated: false + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/20-31-45/.hydra/hydra.yaml b/outputs/2026-06-12/20-31-45/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c44d1a1bdf01dc9230cf986389ba02353bea0dcb --- /dev/null +++ b/outputs/2026-06-12/20-31-45/.hydra/hydra.yaml @@ -0,0 +1,233 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=1100 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=none + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=none,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=1100,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/20-31-45 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/20-31-45/.hydra/overrides.yaml b/outputs/2026-06-12/20-31-45/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3ea991f8fd7f385f8fc1231d04a8448120510272 --- /dev/null +++ b/outputs/2026-06-12/20-31-45/.hydra/overrides.yaml @@ -0,0 +1,75 @@ +- +name=train_dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=1100 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=none +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-12/20-31-45/main.log b/outputs/2026-06-12/20-31-45/main.log new file mode 100644 index 0000000000000000000000000000000000000000..13858e904261245f9a058eb4ff77013098599163 --- /dev/null +++ b/outputs/2026-06-12/20-31-45/main.log @@ -0,0 +1,25 @@ +[2026-06-12 20:32:29,945][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-12 20:32:29,945][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-12 20:32:29,945][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-12 20:32:29,964][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-12 20:32:29,970][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-12 20:32:52,650][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:32:52,650][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:32:52,650][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:32:52,650][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:32:52,650][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:32:55,925][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 20:32:55,925][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 20:32:55,925][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 20:32:55,925][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 20:32:55,925][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 20:31:25,200][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-15 20:31:25,201][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:31:25,201][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-15 20:31:25,201][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:31:25,202][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-15 20:31:25,202][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:31:25,202][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-15 20:31:25,202][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:31:25,204][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-15 20:31:25,205][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-12/20-31-46/.hydra/config.yaml b/outputs/2026-06-12/20-31-46/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..969b5c001b3c5895df04636acb7922a0506e92c3 --- /dev/null +++ b/outputs/2026-06-12/20-31-46/.hydra/config.yaml @@ -0,0 +1,239 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 1100 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: none + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + snr_shifted_memory: + enabled: false + mode: step_offset + offset_steps: 1 + apply_anchor: true + apply_dynamic: false + apply_revisit: true + noise_per_target: true + mark_as_generated: false + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-12/20-31-46/.hydra/hydra.yaml b/outputs/2026-06-12/20-31-46/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e92c1a63e16e25d7088d893888852d5a10976428 --- /dev/null +++ b/outputs/2026-06-12/20-31-46/.hydra/hydra.yaml @@ -0,0 +1,233 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=1100 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=none + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=none,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=1100,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-12/20-31-46 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-12/20-31-46/.hydra/overrides.yaml b/outputs/2026-06-12/20-31-46/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3ea991f8fd7f385f8fc1231d04a8448120510272 --- /dev/null +++ b/outputs/2026-06-12/20-31-46/.hydra/overrides.yaml @@ -0,0 +1,75 @@ +- +name=train_dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=1100 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=none +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-12/20-31-46/main.log b/outputs/2026-06-12/20-31-46/main.log new file mode 100644 index 0000000000000000000000000000000000000000..90743b27d7b605f7523254275ae25e3c0765fb14 --- /dev/null +++ b/outputs/2026-06-12/20-31-46/main.log @@ -0,0 +1,38 @@ +[2026-06-12 20:32:22,623][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-12 20:32:22,956][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-12 20:32:22,956][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-12 20:32:22,956][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-12 20:32:22,956][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-12 20:32:22,957][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-12 20:32:26,261][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 20:32:27,232][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-12 20:32:29,945][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-12 20:32:29,955][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-12 20:32:36,204][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-12 20:32:52,650][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:32:52,650][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-12 20:32:53,913][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 834 M +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 887 K +2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 296 K +3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 296 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +6 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +836 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,271.812 Total estimated model params size (MB) +[2026-06-12 20:32:55,925][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-12 20:32:55,925][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 20:31:25,201][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-15 20:31:25,202][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-15 20:31:25,201][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-15 20:31:25,202][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-13/01-38-54/.hydra/config.yaml b/outputs/2026-06-13/01-38-54/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7600f66abc165be3f3a30297a5a7491e9cf9caa4 --- /dev/null +++ b/outputs/2026-06-13/01-38-54/.hydra/config.yaml @@ -0,0 +1,262 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 750000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 10000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 10000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + training_sampling: + revisit_biased_prob: 0.3 + min_gap: 32 + candidate_starts: 8 + train_negatives: + enabled: true + prob: 0.1 + start_step: 20000 + branches: + - wrong_pose + - time_shuffle + - source_matched_random + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: none + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: spatial_conv_pool + downsample_ratio: 1 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 20000 + ramp_steps: 20000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + snr_shifted_memory: + enabled: true + mode: step_offset + offset_steps: 1 + apply_anchor: true + apply_dynamic: false + apply_revisit: true + noise_per_target: true + mark_as_generated: false + revisit_metric: + enabled: true + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + injection: + memory_hidden_size: 512 + pose_aware: true + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_gate_none_snr_shift_rev1cap_8a100_bs4_global32_750k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_rev1cap_8a100_bs4_global32_750k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-13/01-38-54/.hydra/hydra.yaml b/outputs/2026-06-13/01-38-54/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..404847c914dd4bf6edfb8e8e4285f00e0dd33b42 --- /dev/null +++ b/outputs/2026-06-13/01-38-54/.hydra/hydra.yaml @@ -0,0 +1,249 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_gate_none_snr_shift_rev1cap_8a100_bs4_global32_750k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_rev1cap_8a100_bs4_global32_750k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.injection.pose_aware=true + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=20000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.snr_shifted_memory.enabled=true + - ++algorithm.dememwm.snr_shifted_memory.mode=step_offset + - ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 + - ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true + - ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false + - ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true + - ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true + - ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false + - ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 + - ++algorithm.dememwm.training_sampling.min_gap=32 + - ++algorithm.dememwm.train_negatives.enabled=true + - ++algorithm.dememwm.train_negatives.prob=0.10 + - ++algorithm.dememwm.train_negatives.start_step=20000 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=none + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.revisit.compress.downsample_ratio=1 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=10000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=4 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=10000 + - experiment.training.max_steps=750000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=10000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=20000,++algorithm.dememwm.generated_history_proxy.start_step=20000,++algorithm.dememwm.injection.pose_aware=true,++algorithm.dememwm.revisit.compress.downsample_ratio=1,++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=none,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.snr_shifted_memory.apply_anchor=true,++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false,++algorithm.dememwm.snr_shifted_memory.apply_revisit=true,++algorithm.dememwm.snr_shifted_memory.enabled=true,++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false,++algorithm.dememwm.snr_shifted_memory.mode=step_offset,++algorithm.dememwm.snr_shifted_memory.noise_per_target=true,++algorithm.dememwm.snr_shifted_memory.offset_steps=1,++algorithm.dememwm.train_negatives.enabled=true,++algorithm.dememwm.train_negatives.prob=0.10,++algorithm.dememwm.train_negatives.start_step=20000,++algorithm.dememwm.training_sampling.min_gap=32,++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_gate_none_snr_shift_rev1cap_8a100_bs4_global32_750k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_rev1cap_8a100_bs4_global32_750k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=4,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=750000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=10000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-13/01-38-54 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-13/01-38-54/.hydra/overrides.yaml b/outputs/2026-06-13/01-38-54/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..038f66a14d8d4ea292ed8a55c1a0997b8c913eb3 --- /dev/null +++ b/outputs/2026-06-13/01-38-54/.hydra/overrides.yaml @@ -0,0 +1,91 @@ +- +name=train_dememwm_sep_full_seq_gate_none_snr_shift_rev1cap_8a100_bs4_global32_750k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_rev1cap_8a100_bs4_global32_750k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.injection.pose_aware=true +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=20000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.snr_shifted_memory.enabled=true +- ++algorithm.dememwm.snr_shifted_memory.mode=step_offset +- ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 +- ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true +- ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false +- ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true +- ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true +- ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false +- ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 +- ++algorithm.dememwm.training_sampling.min_gap=32 +- ++algorithm.dememwm.train_negatives.enabled=true +- ++algorithm.dememwm.train_negatives.prob=0.10 +- ++algorithm.dememwm.train_negatives.start_step=20000 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=none +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.revisit.compress.downsample_ratio=1 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=10000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=4 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=10000 +- experiment.training.max_steps=750000 diff --git a/outputs/2026-06-13/01-38-54/main.log b/outputs/2026-06-13/01-38-54/main.log new file mode 100644 index 0000000000000000000000000000000000000000..33e4416fff6513375a2892ddb14af1cc7b7208a7 --- /dev/null +++ b/outputs/2026-06-13/01-38-54/main.log @@ -0,0 +1,116 @@ +[2026-06-13 01:39:56,295][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-13 01:39:56,494][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-13 01:39:56,494][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-13 01:39:56,494][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-13 01:39:56,494][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-13 01:39:56,495][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-13 01:40:00,499][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-13 01:40:01,436][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-13 01:40:03,862][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-13 01:40:03,879][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-13 01:40:03,889][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-13 01:40:03,890][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-13 01:40:03,896][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-13 01:40:03,896][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-13 01:40:03,904][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-13 01:40:10,484][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-13 01:40:25,178][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-13 01:40:25,178][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-13 01:40:25,178][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-13 01:40:25,178][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-13 01:40:25,178][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-13 01:40:25,178][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-13 01:40:25,178][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-13 01:40:26,159][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 834 M +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 887 K +2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 296 K +3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 296 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | dememwm_memory_pose_embed | Sequential | 266 K +6 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +7 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +836 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,273.565 Total estimated model params size (MB) +[2026-06-13 01:40:29,954][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-13 01:40:29,954][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-13 01:40:29,954][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-13 01:40:29,954][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-13 01:40:29,954][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-13 01:40:29,954][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-13 01:40:29,954][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-13 04:41:16,354][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_pairs', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,372][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_revisit_frames', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,373][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,373][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/non_revisit_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,373][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,373][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,373][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_minus_non_revisit_lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,374][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,374][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,374][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_psnr_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,374][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips_gap_8_32', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,375][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,375][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,375][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_psnr_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,375][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips_gap_32_128', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,375][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/num_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,376][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_lpips_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,376][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/self_consistency_psnr_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-13 04:41:16,376][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('revisit/revisit_lpips_gap_128_inf', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-16 01:38:57,046][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-16 01:38:57,046][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-16 01:38:57,046][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-16 01:38:57,046][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-16 01:38:57,046][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-16 01:38:57,047][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-16 01:38:57,047][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-16 01:38:57,047][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-16 01:38:57,047][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-16 01:38:57,047][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-16 01:38:57,045][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-16 01:38:57,046][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-16 01:38:57,048][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-16 01:38:57,048][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-16 01:38:57,050][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-16 01:38:57,050][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-16 01:38:57,050][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-16 01:38:57,050][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-16 01:38:57,050][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-16 01:38:57,051][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-16 01:38:57,051][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-16 01:38:57,051][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-16 01:38:57,051][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-16 01:38:57,051][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-16 01:38:57,051][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-16 01:38:57,051][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-16 01:38:57,051][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-16 01:38:57,051][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-13/01-38-55/.hydra/config.yaml b/outputs/2026-06-13/01-38-55/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7600f66abc165be3f3a30297a5a7491e9cf9caa4 --- /dev/null +++ b/outputs/2026-06-13/01-38-55/.hydra/config.yaml @@ -0,0 +1,262 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 750000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 10000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 10000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + training_sampling: + revisit_biased_prob: 0.3 + min_gap: 32 + candidate_starts: 8 + train_negatives: + enabled: true + prob: 0.1 + start_step: 20000 + branches: + - wrong_pose + - time_shuffle + - source_matched_random + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: none + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: spatial_conv_pool + downsample_ratio: 1 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 20000 + ramp_steps: 20000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + snr_shifted_memory: + enabled: true + mode: step_offset + offset_steps: 1 + apply_anchor: true + apply_dynamic: false + apply_revisit: true + noise_per_target: true + mark_as_generated: false + revisit_metric: + enabled: true + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + injection: + memory_hidden_size: 512 + pose_aware: true + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_gate_none_snr_shift_rev1cap_8a100_bs4_global32_750k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_rev1cap_8a100_bs4_global32_750k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-13/01-38-55/.hydra/hydra.yaml b/outputs/2026-06-13/01-38-55/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..6a77ea7057dd383b54ebe380a66755aa71ebda3f --- /dev/null +++ b/outputs/2026-06-13/01-38-55/.hydra/hydra.yaml @@ -0,0 +1,249 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_gate_none_snr_shift_rev1cap_8a100_bs4_global32_750k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_rev1cap_8a100_bs4_global32_750k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.injection.pose_aware=true + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=20000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.snr_shifted_memory.enabled=true + - ++algorithm.dememwm.snr_shifted_memory.mode=step_offset + - ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 + - ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true + - ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false + - ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true + - ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true + - ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false + - ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 + - ++algorithm.dememwm.training_sampling.min_gap=32 + - ++algorithm.dememwm.train_negatives.enabled=true + - ++algorithm.dememwm.train_negatives.prob=0.10 + - ++algorithm.dememwm.train_negatives.start_step=20000 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=none + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.revisit.compress.downsample_ratio=1 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=10000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=4 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=10000 + - experiment.training.max_steps=750000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=10000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=20000,++algorithm.dememwm.generated_history_proxy.start_step=20000,++algorithm.dememwm.injection.pose_aware=true,++algorithm.dememwm.revisit.compress.downsample_ratio=1,++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=none,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.snr_shifted_memory.apply_anchor=true,++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false,++algorithm.dememwm.snr_shifted_memory.apply_revisit=true,++algorithm.dememwm.snr_shifted_memory.enabled=true,++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false,++algorithm.dememwm.snr_shifted_memory.mode=step_offset,++algorithm.dememwm.snr_shifted_memory.noise_per_target=true,++algorithm.dememwm.snr_shifted_memory.offset_steps=1,++algorithm.dememwm.train_negatives.enabled=true,++algorithm.dememwm.train_negatives.prob=0.10,++algorithm.dememwm.train_negatives.start_step=20000,++algorithm.dememwm.training_sampling.min_gap=32,++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_gate_none_snr_shift_rev1cap_8a100_bs4_global32_750k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_rev1cap_8a100_bs4_global32_750k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=4,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=750000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=10000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-13/01-38-55 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-13/01-38-55/.hydra/overrides.yaml b/outputs/2026-06-13/01-38-55/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..038f66a14d8d4ea292ed8a55c1a0997b8c913eb3 --- /dev/null +++ b/outputs/2026-06-13/01-38-55/.hydra/overrides.yaml @@ -0,0 +1,91 @@ +- +name=train_dememwm_sep_full_seq_gate_none_snr_shift_rev1cap_8a100_bs4_global32_750k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_rev1cap_8a100_bs4_global32_750k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.injection.pose_aware=true +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=20000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.snr_shifted_memory.enabled=true +- ++algorithm.dememwm.snr_shifted_memory.mode=step_offset +- ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 +- ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true +- ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false +- ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true +- ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true +- ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false +- ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 +- ++algorithm.dememwm.training_sampling.min_gap=32 +- ++algorithm.dememwm.train_negatives.enabled=true +- ++algorithm.dememwm.train_negatives.prob=0.10 +- ++algorithm.dememwm.train_negatives.start_step=20000 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=none +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.revisit.compress.downsample_ratio=1 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=10000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=4 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=10000 +- experiment.training.max_steps=750000 diff --git a/outputs/2026-06-13/01-38-55/main.log b/outputs/2026-06-13/01-38-55/main.log new file mode 100644 index 0000000000000000000000000000000000000000..0802811f68dfb661f2005dfb734bfd71e5fa2d37 --- /dev/null +++ b/outputs/2026-06-13/01-38-55/main.log @@ -0,0 +1,7 @@ +[2026-06-13 01:40:03,888][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-13 01:40:25,178][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-13 01:40:29,954][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-16 01:38:57,046][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-16 01:38:57,047][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-16 01:38:57,051][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-16 01:38:57,051][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/18-54-45/.hydra/config.yaml b/outputs/2026-06-15/18-54-45/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8cf23fa199620eb0ba7a0c4a2c845205b684f882 --- /dev/null +++ b/outputs/2026-06-15/18-54-45/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/18-54-45/.hydra/hydra.yaml b/outputs/2026-06-15/18-54-45/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d6489daca5f2280a41c2d9facbe0a40a1e97532a --- /dev/null +++ b/outputs/2026-06-15/18-54-45/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/18-54-45 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/18-54-45/.hydra/overrides.yaml b/outputs/2026-06-15/18-54-45/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4572a3376e9ca2fd0da2e5237fe57a6334cc6596 --- /dev/null +++ b/outputs/2026-06-15/18-54-45/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/18-54-45/main.log b/outputs/2026-06-15/18-54-45/main.log new file mode 100644 index 0000000000000000000000000000000000000000..78652dc24e484ffccbc117d71efba82744e7f978 --- /dev/null +++ b/outputs/2026-06-15/18-54-45/main.log @@ -0,0 +1,5 @@ +[2026-06-15 18:55:40,325][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-15 18:56:01,422][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 18:56:05,241][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 18:54:55,083][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-18 18:54:55,539][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/18-54-46/.hydra/config.yaml b/outputs/2026-06-15/18-54-46/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8cf23fa199620eb0ba7a0c4a2c845205b684f882 --- /dev/null +++ b/outputs/2026-06-15/18-54-46/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/18-54-46/.hydra/hydra.yaml b/outputs/2026-06-15/18-54-46/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..500f46ad7faeeb43beb397debf3dfd94821f641e --- /dev/null +++ b/outputs/2026-06-15/18-54-46/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/18-54-46 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/18-54-46/.hydra/overrides.yaml b/outputs/2026-06-15/18-54-46/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4572a3376e9ca2fd0da2e5237fe57a6334cc6596 --- /dev/null +++ b/outputs/2026-06-15/18-54-46/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/18-54-46/main.log b/outputs/2026-06-15/18-54-46/main.log new file mode 100644 index 0000000000000000000000000000000000000000..0c59c6caf974a12750e4dec4f4e77c66307f3cfb --- /dev/null +++ b/outputs/2026-06-15/18-54-46/main.log @@ -0,0 +1,29 @@ +[2026-06-15 18:55:34,415][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-15 18:55:34,576][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-15 18:55:34,576][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-15 18:55:34,576][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-15 18:55:34,576][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-15 18:55:34,577][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-15 18:55:36,899][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-15 18:55:37,814][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-15 18:55:40,345][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-15 18:55:46,981][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-15 18:56:01,422][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 18:56:02,540][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +1.0 B Trainable params +231 M Non-trainable params +1.3 B Total params +5,107.144 Total estimated model params size (MB) +[2026-06-15 18:56:05,241][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 18:54:55,083][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-18 18:54:55,543][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/18-54-47/.hydra/config.yaml b/outputs/2026-06-15/18-54-47/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8cf23fa199620eb0ba7a0c4a2c845205b684f882 --- /dev/null +++ b/outputs/2026-06-15/18-54-47/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/18-54-47/.hydra/hydra.yaml b/outputs/2026-06-15/18-54-47/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..22f622543b8aa950cd9e19d9508044c0b0ec26cc --- /dev/null +++ b/outputs/2026-06-15/18-54-47/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/18-54-47 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/18-54-47/.hydra/overrides.yaml b/outputs/2026-06-15/18-54-47/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4572a3376e9ca2fd0da2e5237fe57a6334cc6596 --- /dev/null +++ b/outputs/2026-06-15/18-54-47/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/18-54-47/main.log b/outputs/2026-06-15/18-54-47/main.log new file mode 100644 index 0000000000000000000000000000000000000000..086e36b2b6b682eeff96136b83608c073c8d494a --- /dev/null +++ b/outputs/2026-06-15/18-54-47/main.log @@ -0,0 +1,10 @@ +[2026-06-15 18:55:40,325][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-15 18:55:40,326][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-15 18:56:01,422][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 18:56:01,422][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 18:56:05,241][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 18:56:05,241][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 18:54:55,083][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-18 18:54:55,084][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-18 18:54:55,536][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-18 18:54:55,549][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/18-54-48/.hydra/config.yaml b/outputs/2026-06-15/18-54-48/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8cf23fa199620eb0ba7a0c4a2c845205b684f882 --- /dev/null +++ b/outputs/2026-06-15/18-54-48/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/18-54-48/.hydra/hydra.yaml b/outputs/2026-06-15/18-54-48/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cb9eee5c85841d9b8421fdcc0cffbaafd76d7003 --- /dev/null +++ b/outputs/2026-06-15/18-54-48/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/18-54-48 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/18-54-48/.hydra/overrides.yaml b/outputs/2026-06-15/18-54-48/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4572a3376e9ca2fd0da2e5237fe57a6334cc6596 --- /dev/null +++ b/outputs/2026-06-15/18-54-48/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/18-54-48/main.log b/outputs/2026-06-15/18-54-48/main.log new file mode 100644 index 0000000000000000000000000000000000000000..4e614f6a9f22b96266f008be2eb96b598f340b2b --- /dev/null +++ b/outputs/2026-06-15/18-54-48/main.log @@ -0,0 +1,5 @@ +[2026-06-15 18:55:40,322][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-15 18:56:01,422][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 18:56:05,241][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 18:54:55,083][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-18 18:54:55,170][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/18-54-49/.hydra/config.yaml b/outputs/2026-06-15/18-54-49/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8cf23fa199620eb0ba7a0c4a2c845205b684f882 --- /dev/null +++ b/outputs/2026-06-15/18-54-49/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/18-54-49/.hydra/hydra.yaml b/outputs/2026-06-15/18-54-49/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ce4a6e4a135bfa1f2a80f20c9de785e1f463e6b7 --- /dev/null +++ b/outputs/2026-06-15/18-54-49/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/18-54-49 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/18-54-49/.hydra/overrides.yaml b/outputs/2026-06-15/18-54-49/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4572a3376e9ca2fd0da2e5237fe57a6334cc6596 --- /dev/null +++ b/outputs/2026-06-15/18-54-49/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/18-54-49/main.log b/outputs/2026-06-15/18-54-49/main.log new file mode 100644 index 0000000000000000000000000000000000000000..10a2d31176363ecdba195ae2f2fd035a74a2dae0 --- /dev/null +++ b/outputs/2026-06-15/18-54-49/main.log @@ -0,0 +1,5 @@ +[2026-06-15 18:55:40,319][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-15 18:56:01,422][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 18:56:05,241][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 18:54:55,083][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-18 18:54:56,547][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/18-54-50/.hydra/config.yaml b/outputs/2026-06-15/18-54-50/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8cf23fa199620eb0ba7a0c4a2c845205b684f882 --- /dev/null +++ b/outputs/2026-06-15/18-54-50/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/18-54-50/.hydra/hydra.yaml b/outputs/2026-06-15/18-54-50/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..04987ac61bb0f795dab3866ca2199b4b00f61bd3 --- /dev/null +++ b/outputs/2026-06-15/18-54-50/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/18-54-50 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/18-54-50/.hydra/overrides.yaml b/outputs/2026-06-15/18-54-50/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4572a3376e9ca2fd0da2e5237fe57a6334cc6596 --- /dev/null +++ b/outputs/2026-06-15/18-54-50/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/18-54-50/main.log b/outputs/2026-06-15/18-54-50/main.log new file mode 100644 index 0000000000000000000000000000000000000000..f9970fc043398e811154d2915fccd073deae6438 --- /dev/null +++ b/outputs/2026-06-15/18-54-50/main.log @@ -0,0 +1,5 @@ +[2026-06-15 18:55:40,310][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-15 18:56:01,422][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 18:56:05,241][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 18:54:55,083][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-18 18:54:56,536][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/18-54-51/.hydra/config.yaml b/outputs/2026-06-15/18-54-51/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8cf23fa199620eb0ba7a0c4a2c845205b684f882 --- /dev/null +++ b/outputs/2026-06-15/18-54-51/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/18-54-51/.hydra/hydra.yaml b/outputs/2026-06-15/18-54-51/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..32d84637392175f9151371cd42d30f455c806948 --- /dev/null +++ b/outputs/2026-06-15/18-54-51/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/18-54-51 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/18-54-51/.hydra/overrides.yaml b/outputs/2026-06-15/18-54-51/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4572a3376e9ca2fd0da2e5237fe57a6334cc6596 --- /dev/null +++ b/outputs/2026-06-15/18-54-51/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/18-54-51/main.log b/outputs/2026-06-15/18-54-51/main.log new file mode 100644 index 0000000000000000000000000000000000000000..921b1dc9c65de510c6bb3458a1c33f5ac1b10296 --- /dev/null +++ b/outputs/2026-06-15/18-54-51/main.log @@ -0,0 +1,5 @@ +[2026-06-15 18:55:40,302][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-15 18:56:01,422][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 18:56:05,241][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 18:54:55,083][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-18 18:54:56,520][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/18-59-48/.hydra/config.yaml b/outputs/2026-06-15/18-59-48/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ac3697b92d3aaa224d8690827ce376ca37c756e0 --- /dev/null +++ b/outputs/2026-06-15/18-59-48/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 4.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/18-59-48/.hydra/hydra.yaml b/outputs/2026-06-15/18-59-48/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3d619fc106f61669f27b1401d755485797319f13 --- /dev/null +++ b/outputs/2026-06-15/18-59-48/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/18-59-48 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/18-59-48/.hydra/overrides.yaml b/outputs/2026-06-15/18-59-48/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9a9470fcb171b9de7af473034d3f72848dcf0e09 --- /dev/null +++ b/outputs/2026-06-15/18-59-48/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/18-59-48/main.log b/outputs/2026-06-15/18-59-48/main.log new file mode 100644 index 0000000000000000000000000000000000000000..27dfa969ccc0efd5d196fa1dab2e499d8ec1be83 --- /dev/null +++ b/outputs/2026-06-15/18-59-48/main.log @@ -0,0 +1,12 @@ +[2026-06-15 19:00:42,132][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-15 19:00:42,154][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-15 19:01:03,484][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 19:01:03,484][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 19:01:07,645][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 19:01:07,645][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 18:59:55,084][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-18 18:59:55,085][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-18 18:59:55,763][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-18 18:59:55,763][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-18 18:59:55,763][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-18 18:59:55,763][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/18-59-49/.hydra/config.yaml b/outputs/2026-06-15/18-59-49/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ac3697b92d3aaa224d8690827ce376ca37c756e0 --- /dev/null +++ b/outputs/2026-06-15/18-59-49/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 4.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/18-59-49/.hydra/hydra.yaml b/outputs/2026-06-15/18-59-49/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ccf9be1a20e78a9cb8b761f46521e8425b0392e0 --- /dev/null +++ b/outputs/2026-06-15/18-59-49/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/18-59-49 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/18-59-49/.hydra/overrides.yaml b/outputs/2026-06-15/18-59-49/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9a9470fcb171b9de7af473034d3f72848dcf0e09 --- /dev/null +++ b/outputs/2026-06-15/18-59-49/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/18-59-49/main.log b/outputs/2026-06-15/18-59-49/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e5b97f42409c66d44bc7dc6fd50e4b512cbbabb6 --- /dev/null +++ b/outputs/2026-06-15/18-59-49/main.log @@ -0,0 +1,10 @@ +[2026-06-15 19:00:42,135][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-15 19:00:42,140][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-15 19:01:03,484][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 19:01:03,484][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 19:01:07,645][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 19:01:07,645][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 18:59:55,762][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-18 18:59:55,763][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-18 18:59:55,762][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-18 18:59:55,763][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/18-59-51/.hydra/config.yaml b/outputs/2026-06-15/18-59-51/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ac3697b92d3aaa224d8690827ce376ca37c756e0 --- /dev/null +++ b/outputs/2026-06-15/18-59-51/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 4.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/18-59-51/.hydra/hydra.yaml b/outputs/2026-06-15/18-59-51/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5970397648a4d951ed6c427c2b470dc6e90549cd --- /dev/null +++ b/outputs/2026-06-15/18-59-51/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/18-59-51 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/18-59-51/.hydra/overrides.yaml b/outputs/2026-06-15/18-59-51/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9a9470fcb171b9de7af473034d3f72848dcf0e09 --- /dev/null +++ b/outputs/2026-06-15/18-59-51/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/18-59-51/main.log b/outputs/2026-06-15/18-59-51/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e7117f86b163d4e67f49ac6828c46ebcebb46ddc --- /dev/null +++ b/outputs/2026-06-15/18-59-51/main.log @@ -0,0 +1,29 @@ +[2026-06-15 19:00:35,281][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-15 19:00:36,030][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-15 19:00:36,030][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-15 19:00:36,030][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-15 19:00:36,030][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-15 19:00:36,031][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-15 19:00:38,673][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-15 19:00:39,574][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-15 19:00:42,126][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-15 19:00:47,753][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-15 19:01:03,484][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 19:01:05,638][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +1.0 B Trainable params +231 M Non-trainable params +1.3 B Total params +5,107.144 Total estimated model params size (MB) +[2026-06-15 19:01:07,645][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 18:59:55,763][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-18 18:59:55,764][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/18-59-52/.hydra/config.yaml b/outputs/2026-06-15/18-59-52/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ac3697b92d3aaa224d8690827ce376ca37c756e0 --- /dev/null +++ b/outputs/2026-06-15/18-59-52/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 4.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/18-59-52/.hydra/hydra.yaml b/outputs/2026-06-15/18-59-52/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5a627322315cda0a95c201feb284d216a706d3c0 --- /dev/null +++ b/outputs/2026-06-15/18-59-52/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/18-59-52 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/18-59-52/.hydra/overrides.yaml b/outputs/2026-06-15/18-59-52/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9a9470fcb171b9de7af473034d3f72848dcf0e09 --- /dev/null +++ b/outputs/2026-06-15/18-59-52/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/18-59-52/main.log b/outputs/2026-06-15/18-59-52/main.log new file mode 100644 index 0000000000000000000000000000000000000000..dd9452ac66f3289f2bc1349800c09724b8de6476 --- /dev/null +++ b/outputs/2026-06-15/18-59-52/main.log @@ -0,0 +1,7 @@ +[2026-06-15 19:00:42,126][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-15 19:01:03,484][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 19:01:07,645][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 18:59:55,084][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-18 18:59:55,085][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-18 18:59:55,090][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-18 18:59:55,091][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/18-59-53/.hydra/config.yaml b/outputs/2026-06-15/18-59-53/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ac3697b92d3aaa224d8690827ce376ca37c756e0 --- /dev/null +++ b/outputs/2026-06-15/18-59-53/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 4.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/18-59-53/.hydra/hydra.yaml b/outputs/2026-06-15/18-59-53/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..38a6f84f1c5cea13107f15e3b75cc87c6da85cf3 --- /dev/null +++ b/outputs/2026-06-15/18-59-53/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/18-59-53 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/18-59-53/.hydra/overrides.yaml b/outputs/2026-06-15/18-59-53/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9a9470fcb171b9de7af473034d3f72848dcf0e09 --- /dev/null +++ b/outputs/2026-06-15/18-59-53/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/18-59-53/main.log b/outputs/2026-06-15/18-59-53/main.log new file mode 100644 index 0000000000000000000000000000000000000000..051742d540ecbe47cab5f2b5b384161b64aae01a --- /dev/null +++ b/outputs/2026-06-15/18-59-53/main.log @@ -0,0 +1,5 @@ +[2026-06-15 19:00:42,148][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-15 19:01:03,484][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 19:01:07,645][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 18:59:55,763][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-18 18:59:55,765][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/18-59-54/.hydra/config.yaml b/outputs/2026-06-15/18-59-54/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ac3697b92d3aaa224d8690827ce376ca37c756e0 --- /dev/null +++ b/outputs/2026-06-15/18-59-54/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 4.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/18-59-54/.hydra/hydra.yaml b/outputs/2026-06-15/18-59-54/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f70ca6fff3e606dc33c70d461d57d0c6ee88042d --- /dev/null +++ b/outputs/2026-06-15/18-59-54/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/18-59-54 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/18-59-54/.hydra/overrides.yaml b/outputs/2026-06-15/18-59-54/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9a9470fcb171b9de7af473034d3f72848dcf0e09 --- /dev/null +++ b/outputs/2026-06-15/18-59-54/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/18-59-54/main.log b/outputs/2026-06-15/18-59-54/main.log new file mode 100644 index 0000000000000000000000000000000000000000..f516c41b588f466ce6636287a8c8270ea1d4681e --- /dev/null +++ b/outputs/2026-06-15/18-59-54/main.log @@ -0,0 +1,5 @@ +[2026-06-15 19:00:42,153][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-15 19:01:03,484][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 19:01:07,645][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 18:59:55,763][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-18 18:59:55,764][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/20-30-31/.hydra/config.yaml b/outputs/2026-06-15/20-30-31/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5f1434a7d8c57452d4e5434eff3f576e8af06352 --- /dev/null +++ b/outputs/2026-06-15/20-30-31/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 0 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/20-30-31/.hydra/hydra.yaml b/outputs/2026-06-15/20-30-31/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..89fe5db3e4d8ff3f27c00e14326f063f2799fbec --- /dev/null +++ b/outputs/2026-06-15/20-30-31/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=0 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=0,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/20-30-31 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/20-30-31/.hydra/overrides.yaml b/outputs/2026-06-15/20-30-31/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cd100937f0dd08e21d35eafd3b5f1531a2ba3251 --- /dev/null +++ b/outputs/2026-06-15/20-30-31/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=0 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/20-30-31/main.log b/outputs/2026-06-15/20-30-31/main.log new file mode 100644 index 0000000000000000000000000000000000000000..455f835303d18d24b4a715eb15abb1170e92a6d4 --- /dev/null +++ b/outputs/2026-06-15/20-30-31/main.log @@ -0,0 +1,5 @@ +[2026-06-15 20:31:13,328][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-15 20:31:32,964][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 20:31:36,431][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 20:30:27,143][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-18 20:30:27,144][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/20-30-32/.hydra/config.yaml b/outputs/2026-06-15/20-30-32/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5f1434a7d8c57452d4e5434eff3f576e8af06352 --- /dev/null +++ b/outputs/2026-06-15/20-30-32/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 0 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/20-30-32/.hydra/hydra.yaml b/outputs/2026-06-15/20-30-32/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9214e682f5c217507508ee41e56f422d51683216 --- /dev/null +++ b/outputs/2026-06-15/20-30-32/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=0 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=0,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/20-30-32 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/20-30-32/.hydra/overrides.yaml b/outputs/2026-06-15/20-30-32/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cd100937f0dd08e21d35eafd3b5f1531a2ba3251 --- /dev/null +++ b/outputs/2026-06-15/20-30-32/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=0 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/20-30-32/main.log b/outputs/2026-06-15/20-30-32/main.log new file mode 100644 index 0000000000000000000000000000000000000000..d11e282333abeeafc9f173c235ab4eb74edec3ce --- /dev/null +++ b/outputs/2026-06-15/20-30-32/main.log @@ -0,0 +1,20 @@ +[2026-06-15 20:31:12,814][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-15 20:31:12,985][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-15 20:31:13,040][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-15 20:31:13,161][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-15 20:31:32,963][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 20:31:32,964][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 20:31:32,964][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 20:31:32,964][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 20:31:36,431][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 20:31:36,431][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 20:31:36,431][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 20:31:36,431][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 20:30:27,142][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-18 20:30:27,144][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-18 20:30:27,143][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-18 20:30:27,144][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-18 20:30:27,143][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-18 20:30:27,144][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-18 20:30:27,144][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-18 20:30:27,144][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/20-30-33/.hydra/config.yaml b/outputs/2026-06-15/20-30-33/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5f1434a7d8c57452d4e5434eff3f576e8af06352 --- /dev/null +++ b/outputs/2026-06-15/20-30-33/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 0 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/20-30-33/.hydra/hydra.yaml b/outputs/2026-06-15/20-30-33/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ff71af1e7f20cd9f620261eff4567d321cad48c0 --- /dev/null +++ b/outputs/2026-06-15/20-30-33/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=0 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=0,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/20-30-33 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/20-30-33/.hydra/overrides.yaml b/outputs/2026-06-15/20-30-33/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cd100937f0dd08e21d35eafd3b5f1531a2ba3251 --- /dev/null +++ b/outputs/2026-06-15/20-30-33/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=0 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/20-30-33/main.log b/outputs/2026-06-15/20-30-33/main.log new file mode 100644 index 0000000000000000000000000000000000000000..ada057e89673c58693817ae477256192f3843256 --- /dev/null +++ b/outputs/2026-06-15/20-30-33/main.log @@ -0,0 +1,5 @@ +[2026-06-15 20:31:12,918][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-15 20:31:32,964][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 20:31:36,431][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 20:30:27,143][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-18 20:30:27,144][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/20-30-34/.hydra/config.yaml b/outputs/2026-06-15/20-30-34/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5f1434a7d8c57452d4e5434eff3f576e8af06352 --- /dev/null +++ b/outputs/2026-06-15/20-30-34/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 0 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/20-30-34/.hydra/hydra.yaml b/outputs/2026-06-15/20-30-34/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e578a581d5a0783b911e9dff63bebe078331ec97 --- /dev/null +++ b/outputs/2026-06-15/20-30-34/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=0 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=0,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/20-30-34 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/20-30-34/.hydra/overrides.yaml b/outputs/2026-06-15/20-30-34/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cd100937f0dd08e21d35eafd3b5f1531a2ba3251 --- /dev/null +++ b/outputs/2026-06-15/20-30-34/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=0 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/20-30-34/main.log b/outputs/2026-06-15/20-30-34/main.log new file mode 100644 index 0000000000000000000000000000000000000000..2ee9adeaafe1fa3276f89c0b899a8ad1de1faee7 --- /dev/null +++ b/outputs/2026-06-15/20-30-34/main.log @@ -0,0 +1,34 @@ +[2026-06-15 20:31:09,908][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-15 20:31:10,112][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-15 20:31:10,112][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-15 20:31:10,112][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-15 20:31:10,112][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-15 20:31:10,113][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-15 20:31:10,335][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-15 20:31:10,525][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-15 20:31:12,711][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-15 20:31:13,990][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-15 20:31:19,195][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-15 20:31:32,964][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 20:31:32,963][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 20:31:34,433][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +1.0 B Trainable params +231 M Non-trainable params +1.3 B Total params +5,107.144 Total estimated model params size (MB) +[2026-06-15 20:31:36,431][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 20:31:36,431][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 20:30:27,143][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-18 20:30:27,143][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-18 20:30:27,144][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-18 20:30:27,144][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/20-32-34/.hydra/config.yaml b/outputs/2026-06-15/20-32-34/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..00a4819247bb946a4bde3e323b2ace07435b877f --- /dev/null +++ b/outputs/2026-06-15/20-32-34/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/20-32-34/.hydra/hydra.yaml b/outputs/2026-06-15/20-32-34/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d8528b05f1ae37b987d0de492b20b8ab58a3f10c --- /dev/null +++ b/outputs/2026-06-15/20-32-34/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/20-32-34 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/20-32-34/.hydra/overrides.yaml b/outputs/2026-06-15/20-32-34/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c327e3e8696cb770e644c8cdb4761d5660506390 --- /dev/null +++ b/outputs/2026-06-15/20-32-34/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/20-32-34/main.log b/outputs/2026-06-15/20-32-34/main.log new file mode 100644 index 0000000000000000000000000000000000000000..ca441f4e1723eddaeffddea2231a0990c4861405 --- /dev/null +++ b/outputs/2026-06-15/20-32-34/main.log @@ -0,0 +1,10 @@ +[2026-06-15 20:33:23,114][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-15 20:33:23,115][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-15 20:33:42,350][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 20:33:42,350][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 20:33:47,591][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 20:33:47,591][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 20:32:27,490][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-18 20:32:27,491][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-18 20:32:27,491][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-18 20:32:27,492][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/20-32-35/.hydra/config.yaml b/outputs/2026-06-15/20-32-35/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..00a4819247bb946a4bde3e323b2ace07435b877f --- /dev/null +++ b/outputs/2026-06-15/20-32-35/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/20-32-35/.hydra/hydra.yaml b/outputs/2026-06-15/20-32-35/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5e0ed0bb12185ffcdfbaf0333ef97814ed26762e --- /dev/null +++ b/outputs/2026-06-15/20-32-35/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/20-32-35 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/20-32-35/.hydra/overrides.yaml b/outputs/2026-06-15/20-32-35/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c327e3e8696cb770e644c8cdb4761d5660506390 --- /dev/null +++ b/outputs/2026-06-15/20-32-35/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/20-32-35/main.log b/outputs/2026-06-15/20-32-35/main.log new file mode 100644 index 0000000000000000000000000000000000000000..846fb91582c5bd72d40a0a20b9638b39f72d03cf --- /dev/null +++ b/outputs/2026-06-15/20-32-35/main.log @@ -0,0 +1,5 @@ +[2026-06-15 20:33:23,739][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-15 20:33:42,350][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 20:33:47,591][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 20:32:27,491][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-18 20:32:27,492][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/20-32-36/.hydra/config.yaml b/outputs/2026-06-15/20-32-36/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..00a4819247bb946a4bde3e323b2ace07435b877f --- /dev/null +++ b/outputs/2026-06-15/20-32-36/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/20-32-36/.hydra/hydra.yaml b/outputs/2026-06-15/20-32-36/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8a332677b086c6f65020062bacef5bf8bb296a8d --- /dev/null +++ b/outputs/2026-06-15/20-32-36/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/20-32-36 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/20-32-36/.hydra/overrides.yaml b/outputs/2026-06-15/20-32-36/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c327e3e8696cb770e644c8cdb4761d5660506390 --- /dev/null +++ b/outputs/2026-06-15/20-32-36/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/20-32-36/main.log b/outputs/2026-06-15/20-32-36/main.log new file mode 100644 index 0000000000000000000000000000000000000000..c329cb223ef3e3548083f717429678ef74dbbd57 --- /dev/null +++ b/outputs/2026-06-15/20-32-36/main.log @@ -0,0 +1,10 @@ +[2026-06-15 20:33:23,114][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-15 20:33:23,133][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-15 20:33:42,350][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 20:33:42,350][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 20:33:47,591][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-15 20:33:47,591][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 20:32:27,490][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-18 20:32:27,490][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-18 20:32:27,492][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-18 20:32:27,492][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/20-32-37/.hydra/config.yaml b/outputs/2026-06-15/20-32-37/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..00a4819247bb946a4bde3e323b2ace07435b877f --- /dev/null +++ b/outputs/2026-06-15/20-32-37/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/20-32-37/.hydra/hydra.yaml b/outputs/2026-06-15/20-32-37/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..26faedda0f5f9af0b52fb4105b5c65481e53f301 --- /dev/null +++ b/outputs/2026-06-15/20-32-37/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/20-32-37 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/20-32-37/.hydra/overrides.yaml b/outputs/2026-06-15/20-32-37/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c327e3e8696cb770e644c8cdb4761d5660506390 --- /dev/null +++ b/outputs/2026-06-15/20-32-37/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/20-32-37/main.log b/outputs/2026-06-15/20-32-37/main.log new file mode 100644 index 0000000000000000000000000000000000000000..603952866cb5a1c01ef47452f7135316fd4016b5 --- /dev/null +++ b/outputs/2026-06-15/20-32-37/main.log @@ -0,0 +1,5 @@ +[2026-06-15 20:33:23,114][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-15 20:33:42,350][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 20:33:47,591][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 20:32:27,492][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-18 20:32:27,493][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/20-32-38/.hydra/config.yaml b/outputs/2026-06-15/20-32-38/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..00a4819247bb946a4bde3e323b2ace07435b877f --- /dev/null +++ b/outputs/2026-06-15/20-32-38/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/20-32-38/.hydra/hydra.yaml b/outputs/2026-06-15/20-32-38/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d9d11eb0fd771b92d05a7b188eeb4b8b0dec1081 --- /dev/null +++ b/outputs/2026-06-15/20-32-38/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/20-32-38 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/20-32-38/.hydra/overrides.yaml b/outputs/2026-06-15/20-32-38/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c327e3e8696cb770e644c8cdb4761d5660506390 --- /dev/null +++ b/outputs/2026-06-15/20-32-38/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/20-32-38/main.log b/outputs/2026-06-15/20-32-38/main.log new file mode 100644 index 0000000000000000000000000000000000000000..b78e91d5d7b9b220f44e049ce497f04270fe7f0f --- /dev/null +++ b/outputs/2026-06-15/20-32-38/main.log @@ -0,0 +1,29 @@ +[2026-06-15 20:33:16,765][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-15 20:33:16,898][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-15 20:33:16,898][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-15 20:33:16,898][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-15 20:33:16,898][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-15 20:33:16,899][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-15 20:33:19,199][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-15 20:33:20,394][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-15 20:33:23,139][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-15 20:33:29,340][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-15 20:33:42,350][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 20:33:43,567][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +1.0 B Trainable params +231 M Non-trainable params +1.3 B Total params +5,107.144 Total estimated model params size (MB) +[2026-06-15 20:33:47,591][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 20:32:27,492][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-18 20:32:27,493][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-15/20-32-39/.hydra/config.yaml b/outputs/2026-06-15/20-32-39/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..00a4819247bb946a4bde3e323b2ace07435b877f --- /dev/null +++ b/outputs/2026-06-15/20-32-39/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-15/20-32-39/.hydra/hydra.yaml b/outputs/2026-06-15/20-32-39/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..198cc560030c04bc59cd5521e30748ce75993f01 --- /dev/null +++ b/outputs/2026-06-15/20-32-39/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-15/20-32-39 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-15/20-32-39/.hydra/overrides.yaml b/outputs/2026-06-15/20-32-39/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c327e3e8696cb770e644c8cdb4761d5660506390 --- /dev/null +++ b/outputs/2026-06-15/20-32-39/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-15/20-32-39/main.log b/outputs/2026-06-15/20-32-39/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e2e3746ad1c7856d42130c22d61fea2b1d257308 --- /dev/null +++ b/outputs/2026-06-15/20-32-39/main.log @@ -0,0 +1,5 @@ +[2026-06-15 20:33:23,115][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-15 20:33:42,350][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-15 20:33:47,591][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-18 20:32:27,491][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-18 20:32:27,492][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-16/00-48-19/.hydra/config.yaml b/outputs/2026-06-16/00-48-19/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9869f51c161adbbd6d3c059d1e67ed8908232dc6 --- /dev/null +++ b/outputs/2026-06-16/00-48-19/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-16/00-48-19/.hydra/hydra.yaml b/outputs/2026-06-16/00-48-19/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..aa5c29d08e1b53e2bbffe68441bf2dcfe31ba9b7 --- /dev/null +++ b/outputs/2026-06-16/00-48-19/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-16/00-48-19 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-16/00-48-19/.hydra/overrides.yaml b/outputs/2026-06-16/00-48-19/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5e998fce7e79fdbf9bdb2057d2f99d7858591ce3 --- /dev/null +++ b/outputs/2026-06-16/00-48-19/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-16/00-48-19/main.log b/outputs/2026-06-16/00-48-19/main.log new file mode 100644 index 0000000000000000000000000000000000000000..b74c63b1be0667da941566c665bb39f9ca92ad4f --- /dev/null +++ b/outputs/2026-06-16/00-48-19/main.log @@ -0,0 +1,5 @@ +[2026-06-16 00:49:13,101][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-16 00:49:33,682][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-16 00:49:39,817][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 00:48:29,080][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-19 00:48:30,247][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-16/00-48-21/.hydra/config.yaml b/outputs/2026-06-16/00-48-21/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9869f51c161adbbd6d3c059d1e67ed8908232dc6 --- /dev/null +++ b/outputs/2026-06-16/00-48-21/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-16/00-48-21/.hydra/hydra.yaml b/outputs/2026-06-16/00-48-21/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8cfe2c304c28c836df53aac2beb0b83b34636238 --- /dev/null +++ b/outputs/2026-06-16/00-48-21/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-16/00-48-21 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-16/00-48-21/.hydra/overrides.yaml b/outputs/2026-06-16/00-48-21/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5e998fce7e79fdbf9bdb2057d2f99d7858591ce3 --- /dev/null +++ b/outputs/2026-06-16/00-48-21/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-16/00-48-21/main.log b/outputs/2026-06-16/00-48-21/main.log new file mode 100644 index 0000000000000000000000000000000000000000..af31a22113a1c1ed595ea2a8b7ddd242112deb03 --- /dev/null +++ b/outputs/2026-06-16/00-48-21/main.log @@ -0,0 +1,5 @@ +[2026-06-16 00:49:13,100][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-16 00:49:33,682][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-16 00:49:39,817][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 00:48:29,079][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-19 00:48:30,234][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-16/00-48-22/.hydra/config.yaml b/outputs/2026-06-16/00-48-22/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9869f51c161adbbd6d3c059d1e67ed8908232dc6 --- /dev/null +++ b/outputs/2026-06-16/00-48-22/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-16/00-48-22/.hydra/hydra.yaml b/outputs/2026-06-16/00-48-22/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7bfe670eafac9ab2023648587771f36c401ae420 --- /dev/null +++ b/outputs/2026-06-16/00-48-22/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-16/00-48-22 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-16/00-48-22/.hydra/overrides.yaml b/outputs/2026-06-16/00-48-22/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5e998fce7e79fdbf9bdb2057d2f99d7858591ce3 --- /dev/null +++ b/outputs/2026-06-16/00-48-22/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-16/00-48-22/main.log b/outputs/2026-06-16/00-48-22/main.log new file mode 100644 index 0000000000000000000000000000000000000000..ba34ddaff4120898601f6183c68c2099cf3e8c44 --- /dev/null +++ b/outputs/2026-06-16/00-48-22/main.log @@ -0,0 +1,34 @@ +[2026-06-16 00:49:06,036][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-16 00:49:06,170][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-16 00:49:06,171][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-16 00:49:06,171][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-16 00:49:06,171][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-16 00:49:06,172][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-16 00:49:09,011][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-16 00:49:10,152][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-16 00:49:13,099][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-16 00:49:13,109][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-16 00:49:18,823][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-16 00:49:33,682][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-16 00:49:33,682][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-16 00:49:35,840][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +1.0 B Trainable params +231 M Non-trainable params +1.3 B Total params +5,107.144 Total estimated model params size (MB) +[2026-06-16 00:49:39,817][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-16 00:49:39,817][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 00:48:29,079][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-19 00:48:29,080][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-19 00:48:29,168][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-19 00:48:29,259][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-16/00-48-23/.hydra/config.yaml b/outputs/2026-06-16/00-48-23/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9869f51c161adbbd6d3c059d1e67ed8908232dc6 --- /dev/null +++ b/outputs/2026-06-16/00-48-23/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-16/00-48-23/.hydra/hydra.yaml b/outputs/2026-06-16/00-48-23/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9201ccd74144c5924f8aff31bd6879699bac5a7b --- /dev/null +++ b/outputs/2026-06-16/00-48-23/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-16/00-48-23 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-16/00-48-23/.hydra/overrides.yaml b/outputs/2026-06-16/00-48-23/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5e998fce7e79fdbf9bdb2057d2f99d7858591ce3 --- /dev/null +++ b/outputs/2026-06-16/00-48-23/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-16/00-48-23/main.log b/outputs/2026-06-16/00-48-23/main.log new file mode 100644 index 0000000000000000000000000000000000000000..188f3becabb2d2f72d3b3b6ab29e670bedbee0a1 --- /dev/null +++ b/outputs/2026-06-16/00-48-23/main.log @@ -0,0 +1,10 @@ +[2026-06-16 00:49:13,100][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-16 00:49:13,127][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-16 00:49:33,682][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-16 00:49:33,682][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-16 00:49:39,817][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-16 00:49:39,817][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 00:48:29,080][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-19 00:48:29,080][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-19 00:48:30,236][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-19 00:48:30,255][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-16/00-48-24/.hydra/config.yaml b/outputs/2026-06-16/00-48-24/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9869f51c161adbbd6d3c059d1e67ed8908232dc6 --- /dev/null +++ b/outputs/2026-06-16/00-48-24/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-16/00-48-24/.hydra/hydra.yaml b/outputs/2026-06-16/00-48-24/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4de5e6405e5b0cbdf39fe514789e015370235a63 --- /dev/null +++ b/outputs/2026-06-16/00-48-24/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-16/00-48-24 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-16/00-48-24/.hydra/overrides.yaml b/outputs/2026-06-16/00-48-24/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5e998fce7e79fdbf9bdb2057d2f99d7858591ce3 --- /dev/null +++ b/outputs/2026-06-16/00-48-24/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-16/00-48-24/main.log b/outputs/2026-06-16/00-48-24/main.log new file mode 100644 index 0000000000000000000000000000000000000000..02999ed3bd877431fb8da996ce0b0fe874c9a2a2 --- /dev/null +++ b/outputs/2026-06-16/00-48-24/main.log @@ -0,0 +1,5 @@ +[2026-06-16 00:49:13,098][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-16 00:49:33,682][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-16 00:49:39,817][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 00:48:29,080][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-19 00:48:30,249][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-16/00-48-25/.hydra/config.yaml b/outputs/2026-06-16/00-48-25/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9869f51c161adbbd6d3c059d1e67ed8908232dc6 --- /dev/null +++ b/outputs/2026-06-16/00-48-25/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-16/00-48-25/.hydra/hydra.yaml b/outputs/2026-06-16/00-48-25/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fd5290bdfc8f05d6b5907aa817d27f8b73c3cce5 --- /dev/null +++ b/outputs/2026-06-16/00-48-25/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-16/00-48-25 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-16/00-48-25/.hydra/overrides.yaml b/outputs/2026-06-16/00-48-25/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5e998fce7e79fdbf9bdb2057d2f99d7858591ce3 --- /dev/null +++ b/outputs/2026-06-16/00-48-25/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-16/00-48-25/main.log b/outputs/2026-06-16/00-48-25/main.log new file mode 100644 index 0000000000000000000000000000000000000000..44e4ec1b3beb723704a2f58dc95f3e136a2d5131 --- /dev/null +++ b/outputs/2026-06-16/00-48-25/main.log @@ -0,0 +1,5 @@ +[2026-06-16 00:49:13,125][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-16 00:49:33,682][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-16 00:49:39,817][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 00:48:29,080][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-19 00:48:29,245][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-16/01-22-05/.hydra/config.yaml b/outputs/2026-06-16/01-22-05/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d448e9f3d5b0f919cdd9db5960181d659ab5f07c --- /dev/null +++ b/outputs/2026-06-16/01-22-05/.hydra/config.yaml @@ -0,0 +1,262 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 750000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 10000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 10000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + training_sampling: + revisit_biased_prob: 0.3 + min_gap: 32 + candidate_starts: 8 + train_negatives: + enabled: true + prob: 0.1 + start_step: 20000 + branches: + - wrong_pose + - time_shuffle + - source_matched_random + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: none + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 20000 + ramp_steps: 20000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + snr_shifted_memory: + enabled: true + mode: step_offset + offset_steps: 1 + apply_anchor: true + apply_dynamic: false + apply_revisit: true + noise_per_target: true + mark_as_generated: false + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + injection: + memory_hidden_size: 512 + pose_aware: true + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-16/01-22-05/.hydra/hydra.yaml b/outputs/2026-06-16/01-22-05/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fdb58455d25e31211f6884ce8fc19a4295bd6c77 --- /dev/null +++ b/outputs/2026-06-16/01-22-05/.hydra/hydra.yaml @@ -0,0 +1,250 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.injection.pose_aware=true + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=20000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.snr_shifted_memory.enabled=true + - ++algorithm.dememwm.snr_shifted_memory.mode=step_offset + - ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 + - ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true + - ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false + - ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true + - ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true + - ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false + - ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 + - ++algorithm.dememwm.training_sampling.min_gap=32 + - ++algorithm.dememwm.train_negatives.enabled=true + - ++algorithm.dememwm.train_negatives.prob=0.10 + - ++algorithm.dememwm.train_negatives.start_step=20000 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=none + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=10000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=4 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=10000 + - experiment.training.max_steps=750000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=10000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=20000,++algorithm.dememwm.generated_history_proxy.start_step=20000,++algorithm.dememwm.injection.pose_aware=true,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=none,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.dememwm.snr_shifted_memory.apply_anchor=true,++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false,++algorithm.dememwm.snr_shifted_memory.apply_revisit=true,++algorithm.dememwm.snr_shifted_memory.enabled=true,++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false,++algorithm.dememwm.snr_shifted_memory.mode=step_offset,++algorithm.dememwm.snr_shifted_memory.noise_per_target=true,++algorithm.dememwm.snr_shifted_memory.offset_steps=1,++algorithm.dememwm.train_negatives.enabled=true,++algorithm.dememwm.train_negatives.prob=0.10,++algorithm.dememwm.train_negatives.start_step=20000,++algorithm.dememwm.training_sampling.min_gap=32,++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=4,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=750000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=10000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-16/01-22-05 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-16/01-22-05/.hydra/overrides.yaml b/outputs/2026-06-16/01-22-05/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..de99b40af9c811edc833b8fa7d40eac9948a8baa --- /dev/null +++ b/outputs/2026-06-16/01-22-05/.hydra/overrides.yaml @@ -0,0 +1,92 @@ +- +name=train_dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.injection.pose_aware=true +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=20000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.snr_shifted_memory.enabled=true +- ++algorithm.dememwm.snr_shifted_memory.mode=step_offset +- ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 +- ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true +- ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false +- ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true +- ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true +- ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false +- ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 +- ++algorithm.dememwm.training_sampling.min_gap=32 +- ++algorithm.dememwm.train_negatives.enabled=true +- ++algorithm.dememwm.train_negatives.prob=0.10 +- ++algorithm.dememwm.train_negatives.start_step=20000 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=none +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=10000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=4 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=10000 +- experiment.training.max_steps=750000 diff --git a/outputs/2026-06-16/01-22-05/main.log b/outputs/2026-06-16/01-22-05/main.log new file mode 100644 index 0000000000000000000000000000000000000000..621c56d94e3f06f89a6238bf82b69921141b63b4 --- /dev/null +++ b/outputs/2026-06-16/01-22-05/main.log @@ -0,0 +1,15 @@ +[2026-06-16 01:22:51,951][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-16 01:22:52,671][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-16 01:22:54,087][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-16 01:23:35,952][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-16 01:23:35,985][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-16 01:23:36,330][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-16 01:23:41,941][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-16 01:23:41,941][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-16 01:23:41,941][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 01:21:59,077][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-19 01:21:59,078][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-19 01:21:59,078][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-19 01:22:00,081][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-19 01:22:01,013][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-19 01:22:01,993][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-16/01-22-06/.hydra/config.yaml b/outputs/2026-06-16/01-22-06/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d448e9f3d5b0f919cdd9db5960181d659ab5f07c --- /dev/null +++ b/outputs/2026-06-16/01-22-06/.hydra/config.yaml @@ -0,0 +1,262 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 750000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 10000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: null + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 10000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + training_sampling: + revisit_biased_prob: 0.3 + min_gap: 32 + candidate_starts: 8 + train_negatives: + enabled: true + prob: 0.1 + start_step: 20000 + branches: + - wrong_pose + - time_shuffle + - source_matched_random + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + gate_mode: none + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + tokenization: spatial_conv_pool + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 20000 + ramp_steps: 20000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + snr_shifted_memory: + enabled: true + mode: step_offset + offset_steps: 1 + apply_anchor: true + apply_dynamic: false + apply_revisit: true + noise_per_target: true + mark_as_generated: false + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + injection: + memory_hidden_size: 512 + pose_aware: true + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-16/01-22-06/.hydra/hydra.yaml b/outputs/2026-06-16/01-22-06/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..6da994d16f159663689d367c40724a6d68dc498c --- /dev/null +++ b/outputs/2026-06-16/01-22-06/.hydra/hydra.yaml @@ -0,0 +1,250 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.injection.pose_aware=true + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=20000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.snr_shifted_memory.enabled=true + - ++algorithm.dememwm.snr_shifted_memory.mode=step_offset + - ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 + - ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true + - ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false + - ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true + - ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true + - ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false + - ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 + - ++algorithm.dememwm.training_sampling.min_gap=32 + - ++algorithm.dememwm.train_negatives.enabled=true + - ++algorithm.dememwm.train_negatives.prob=0.10 + - ++algorithm.dememwm.train_negatives.start_step=20000 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.gate_mode=none + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=10000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=4 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=10000 + - experiment.training.max_steps=750000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=10000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=20000,++algorithm.dememwm.generated_history_proxy.start_step=20000,++algorithm.dememwm.injection.pose_aware=true,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.gate_mode=none,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.dememwm.snr_shifted_memory.apply_anchor=true,++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false,++algorithm.dememwm.snr_shifted_memory.apply_revisit=true,++algorithm.dememwm.snr_shifted_memory.enabled=true,++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false,++algorithm.dememwm.snr_shifted_memory.mode=step_offset,++algorithm.dememwm.snr_shifted_memory.noise_per_target=true,++algorithm.dememwm.snr_shifted_memory.offset_steps=1,++algorithm.dememwm.train_negatives.enabled=true,++algorithm.dememwm.train_negatives.prob=0.10,++algorithm.dememwm.train_negatives.start_step=20000,++algorithm.dememwm.training_sampling.min_gap=32,++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=4,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=750000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=10000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-16/01-22-06 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-16/01-22-06/.hydra/overrides.yaml b/outputs/2026-06-16/01-22-06/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..de99b40af9c811edc833b8fa7d40eac9948a8baa --- /dev/null +++ b/outputs/2026-06-16/01-22-06/.hydra/overrides.yaml @@ -0,0 +1,92 @@ +- +name=train_dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.injection.pose_aware=true +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=20000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=20000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.snr_shifted_memory.enabled=true +- ++algorithm.dememwm.snr_shifted_memory.mode=step_offset +- ++algorithm.dememwm.snr_shifted_memory.offset_steps=1 +- ++algorithm.dememwm.snr_shifted_memory.apply_anchor=true +- ++algorithm.dememwm.snr_shifted_memory.apply_dynamic=false +- ++algorithm.dememwm.snr_shifted_memory.apply_revisit=true +- ++algorithm.dememwm.snr_shifted_memory.noise_per_target=true +- ++algorithm.dememwm.snr_shifted_memory.mark_as_generated=false +- ++algorithm.dememwm.training_sampling.revisit_biased_prob=0.3 +- ++algorithm.dememwm.training_sampling.min_gap=32 +- ++algorithm.dememwm.train_negatives.enabled=true +- ++algorithm.dememwm.train_negatives.prob=0.10 +- ++algorithm.dememwm.train_negatives.start_step=20000 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.gate_mode=none +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.tokenization=spatial_conv_pool +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=10000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=4 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=10000 +- experiment.training.max_steps=750000 diff --git a/outputs/2026-06-16/01-22-06/main.log b/outputs/2026-06-16/01-22-06/main.log new file mode 100644 index 0000000000000000000000000000000000000000..31646a7abc628c0749db0033787e08840872ebd9 --- /dev/null +++ b/outputs/2026-06-16/01-22-06/main.log @@ -0,0 +1,54 @@ +[2026-06-16 01:22:48,879][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-16 01:22:49,061][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-16 01:22:49,061][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-16 01:22:49,061][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-16 01:22:49,061][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-16 01:22:49,061][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-16 01:22:51,816][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-16 01:22:51,823][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-16 01:22:51,942][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-16 01:22:51,947][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-16 01:22:52,595][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-16 01:22:58,817][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-16 01:23:13,076][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/checkpoints/last.ckpt +[2026-06-16 01:23:35,880][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-16 01:23:36,308][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-16 01:23:36,308][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-16 01:23:36,329][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-16 01:23:36,352][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-16 01:23:39,909][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 834 M +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 887 K +2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 296 K +3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 296 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | dememwm_memory_pose_embed | Sequential | 266 K +6 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +7 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +836 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,273.565 Total estimated model params size (MB) +[2026-06-16 01:23:40,718][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_sep_full_seq_gate_none_snr_shift_ntok8_8a100_bs4_global32_750k/checkpoints/last.ckpt +[2026-06-16 01:23:41,941][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-16 01:23:41,941][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-16 01:23:41,941][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-16 01:23:41,941][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-16 01:23:41,941][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 01:21:59,077][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-19 01:21:59,077][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-19 01:21:59,078][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-19 01:21:59,078][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-19 01:21:59,078][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-19 01:21:59,154][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-19 01:21:59,155][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-19 01:22:00,988][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-19 01:22:01,006][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-19 01:22:02,010][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/06-56-28/.hydra/config.yaml b/outputs/2026-06-19/06-56-28/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8cf23fa199620eb0ba7a0c4a2c845205b684f882 --- /dev/null +++ b/outputs/2026-06-19/06-56-28/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/06-56-28/.hydra/hydra.yaml b/outputs/2026-06-19/06-56-28/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c6a3c838718da06c801285736889ea28d017a2ea --- /dev/null +++ b/outputs/2026-06-19/06-56-28/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/06-56-28 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/06-56-28/.hydra/overrides.yaml b/outputs/2026-06-19/06-56-28/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4572a3376e9ca2fd0da2e5237fe57a6334cc6596 --- /dev/null +++ b/outputs/2026-06-19/06-56-28/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/06-56-28/main.log b/outputs/2026-06-19/06-56-28/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e6bc5345ee58f21825c11809f7f4d9733a6108c4 --- /dev/null +++ b/outputs/2026-06-19/06-56-28/main.log @@ -0,0 +1,39 @@ +[2026-06-19 06:57:13,669][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-19 06:57:13,810][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-19 06:57:13,810][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-19 06:57:13,810][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-19 06:57:13,810][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-19 06:57:13,811][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-19 06:57:16,523][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-19 06:57:16,524][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-19 06:57:17,008][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-19 06:57:23,488][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-19 06:57:38,413][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/checkpoints/last.ckpt +[2026-06-19 06:58:01,475][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 06:58:02,077][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 06:58:02,100][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 06:58:03,288][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +1.0 B Trainable params +231 M Non-trainable params +1.3 B Total params +5,107.144 Total estimated model params size (MB) +[2026-06-19 06:58:04,450][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/checkpoints/last.ckpt +[2026-06-19 06:58:06,695][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 06:58:06,695][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 06:58:06,696][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 06:56:31,692][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-22 06:56:31,692][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-22 06:56:31,693][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-22 06:56:31,693][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-22 06:56:31,693][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-22 06:56:31,693][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/06-56-29/.hydra/config.yaml b/outputs/2026-06-19/06-56-29/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8cf23fa199620eb0ba7a0c4a2c845205b684f882 --- /dev/null +++ b/outputs/2026-06-19/06-56-29/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/06-56-29/.hydra/hydra.yaml b/outputs/2026-06-19/06-56-29/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d48b394c02e701f93159597334b974e52b9d3012 --- /dev/null +++ b/outputs/2026-06-19/06-56-29/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/06-56-29 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/06-56-29/.hydra/overrides.yaml b/outputs/2026-06-19/06-56-29/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4572a3376e9ca2fd0da2e5237fe57a6334cc6596 --- /dev/null +++ b/outputs/2026-06-19/06-56-29/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/06-56-29/main.log b/outputs/2026-06-19/06-56-29/main.log new file mode 100644 index 0000000000000000000000000000000000000000..7c30e9db4b34b2e227a84a44171068e555f3461e --- /dev/null +++ b/outputs/2026-06-19/06-56-29/main.log @@ -0,0 +1,10 @@ +[2026-06-19 06:57:16,525][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-19 06:57:16,569][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-19 06:58:01,446][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 06:58:02,170][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 06:58:06,695][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 06:58:06,695][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 06:56:31,693][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-22 06:56:31,694][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-22 06:56:31,693][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-22 06:56:31,696][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/06-56-30/.hydra/config.yaml b/outputs/2026-06-19/06-56-30/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8cf23fa199620eb0ba7a0c4a2c845205b684f882 --- /dev/null +++ b/outputs/2026-06-19/06-56-30/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/06-56-30/.hydra/hydra.yaml b/outputs/2026-06-19/06-56-30/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..1ed6bfc555a6678cf3bbd6d204eb9b2ee181323f --- /dev/null +++ b/outputs/2026-06-19/06-56-30/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/06-56-30 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/06-56-30/.hydra/overrides.yaml b/outputs/2026-06-19/06-56-30/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4572a3376e9ca2fd0da2e5237fe57a6334cc6596 --- /dev/null +++ b/outputs/2026-06-19/06-56-30/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/06-56-30/main.log b/outputs/2026-06-19/06-56-30/main.log new file mode 100644 index 0000000000000000000000000000000000000000..6330f1c22293ea31debf9e836ae6fd6b47e3be6a --- /dev/null +++ b/outputs/2026-06-19/06-56-30/main.log @@ -0,0 +1,5 @@ +[2026-06-19 06:57:16,556][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-19 06:58:01,470][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 06:58:06,696][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 06:56:31,692][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-22 06:56:31,693][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/06-56-31/.hydra/config.yaml b/outputs/2026-06-19/06-56-31/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8cf23fa199620eb0ba7a0c4a2c845205b684f882 --- /dev/null +++ b/outputs/2026-06-19/06-56-31/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/06-56-31/.hydra/hydra.yaml b/outputs/2026-06-19/06-56-31/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b8c10ea3a879eb527c6aa94052e01d67de4d4fd1 --- /dev/null +++ b/outputs/2026-06-19/06-56-31/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/06-56-31 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/06-56-31/.hydra/overrides.yaml b/outputs/2026-06-19/06-56-31/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4572a3376e9ca2fd0da2e5237fe57a6334cc6596 --- /dev/null +++ b/outputs/2026-06-19/06-56-31/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/06-56-31/main.log b/outputs/2026-06-19/06-56-31/main.log new file mode 100644 index 0000000000000000000000000000000000000000..4ea7a7c87651a826aebb7a9cb0ec0dccb591a535 --- /dev/null +++ b/outputs/2026-06-19/06-56-31/main.log @@ -0,0 +1,5 @@ +[2026-06-19 06:57:16,563][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-19 06:58:02,103][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 06:58:06,695][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 06:56:31,692][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-22 06:56:31,693][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/06-56-32/.hydra/config.yaml b/outputs/2026-06-19/06-56-32/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8cf23fa199620eb0ba7a0c4a2c845205b684f882 --- /dev/null +++ b/outputs/2026-06-19/06-56-32/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/06-56-32/.hydra/hydra.yaml b/outputs/2026-06-19/06-56-32/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9170577acac6ee1ff1c8fea984915c226f173063 --- /dev/null +++ b/outputs/2026-06-19/06-56-32/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/06-56-32 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/06-56-32/.hydra/overrides.yaml b/outputs/2026-06-19/06-56-32/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4572a3376e9ca2fd0da2e5237fe57a6334cc6596 --- /dev/null +++ b/outputs/2026-06-19/06-56-32/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/06-56-32/main.log b/outputs/2026-06-19/06-56-32/main.log new file mode 100644 index 0000000000000000000000000000000000000000..52ecd14b522d7c50c809a87903a44d796957785c --- /dev/null +++ b/outputs/2026-06-19/06-56-32/main.log @@ -0,0 +1,5 @@ +[2026-06-19 06:57:16,521][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-19 06:58:02,149][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 06:58:06,695][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 06:56:31,693][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-22 06:56:31,694][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/12-46-30/.hydra/config.yaml b/outputs/2026-06-19/12-46-30/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ac3697b92d3aaa224d8690827ce376ca37c756e0 --- /dev/null +++ b/outputs/2026-06-19/12-46-30/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 4.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/12-46-30/.hydra/hydra.yaml b/outputs/2026-06-19/12-46-30/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..bb3bfc3be7ccb12eaa19678247b6d55de940be4f --- /dev/null +++ b/outputs/2026-06-19/12-46-30/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/12-46-30 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/12-46-30/.hydra/overrides.yaml b/outputs/2026-06-19/12-46-30/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9a9470fcb171b9de7af473034d3f72848dcf0e09 --- /dev/null +++ b/outputs/2026-06-19/12-46-30/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/12-46-30/main.log b/outputs/2026-06-19/12-46-30/main.log new file mode 100644 index 0000000000000000000000000000000000000000..d2e134a4e16fd2d510e81708eca25eb4cf62a5cd --- /dev/null +++ b/outputs/2026-06-19/12-46-30/main.log @@ -0,0 +1,10 @@ +[2026-06-19 12:47:21,735][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-19 12:47:21,814][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-19 12:48:13,581][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 12:48:13,598][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 12:48:17,646][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 12:48:17,646][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 12:46:33,051][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-22 12:46:33,051][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-22 12:46:34,526][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-22 12:46:35,544][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/12-46-31/.hydra/config.yaml b/outputs/2026-06-19/12-46-31/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ac3697b92d3aaa224d8690827ce376ca37c756e0 --- /dev/null +++ b/outputs/2026-06-19/12-46-31/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 4.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/12-46-31/.hydra/hydra.yaml b/outputs/2026-06-19/12-46-31/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8e47ff562981b481cb496c37149af887201c547e --- /dev/null +++ b/outputs/2026-06-19/12-46-31/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/12-46-31 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/12-46-31/.hydra/overrides.yaml b/outputs/2026-06-19/12-46-31/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9a9470fcb171b9de7af473034d3f72848dcf0e09 --- /dev/null +++ b/outputs/2026-06-19/12-46-31/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/12-46-31/main.log b/outputs/2026-06-19/12-46-31/main.log new file mode 100644 index 0000000000000000000000000000000000000000..3b2e60b442124d19ba5052ede2555d1c480b5d43 --- /dev/null +++ b/outputs/2026-06-19/12-46-31/main.log @@ -0,0 +1,39 @@ +[2026-06-19 12:47:18,197][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-19 12:47:21,157][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-19 12:47:21,284][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-19 12:47:21,284][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-19 12:47:21,284][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-19 12:47:21,285][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-19 12:47:21,285][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-19 12:47:22,686][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-19 12:47:23,441][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-19 12:47:29,445][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-19 12:47:45,097][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/checkpoints/last.ckpt +[2026-06-19 12:48:12,784][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 12:48:12,797][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 12:48:13,619][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 12:48:15,122][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +1.0 B Trainable params +231 M Non-trainable params +1.3 B Total params +5,107.144 Total estimated model params size (MB) +[2026-06-19 12:48:16,709][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/checkpoints/last.ckpt +[2026-06-19 12:48:17,646][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 12:48:17,646][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 12:48:17,646][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 12:46:33,051][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-22 12:46:33,051][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-22 12:46:33,052][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-22 12:46:33,145][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-22 12:46:33,506][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-22 12:46:33,540][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/12-46-32/.hydra/config.yaml b/outputs/2026-06-19/12-46-32/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ac3697b92d3aaa224d8690827ce376ca37c756e0 --- /dev/null +++ b/outputs/2026-06-19/12-46-32/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 4.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/12-46-32/.hydra/hydra.yaml b/outputs/2026-06-19/12-46-32/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7bb0498eff38ac18041ffbe05e58ee626b1e3406 --- /dev/null +++ b/outputs/2026-06-19/12-46-32/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/12-46-32 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/12-46-32/.hydra/overrides.yaml b/outputs/2026-06-19/12-46-32/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9a9470fcb171b9de7af473034d3f72848dcf0e09 --- /dev/null +++ b/outputs/2026-06-19/12-46-32/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/12-46-32/main.log b/outputs/2026-06-19/12-46-32/main.log new file mode 100644 index 0000000000000000000000000000000000000000..c15367432493af7819abc6f63cfa6a74b62f5228 --- /dev/null +++ b/outputs/2026-06-19/12-46-32/main.log @@ -0,0 +1,5 @@ +[2026-06-19 12:47:21,437][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-19 12:48:12,771][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 12:48:17,646][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 12:46:33,051][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-22 12:46:34,527][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/12-46-33/.hydra/config.yaml b/outputs/2026-06-19/12-46-33/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ac3697b92d3aaa224d8690827ce376ca37c756e0 --- /dev/null +++ b/outputs/2026-06-19/12-46-33/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 4.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/12-46-33/.hydra/hydra.yaml b/outputs/2026-06-19/12-46-33/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fc48166484743a940ead651c521e67494962dcad --- /dev/null +++ b/outputs/2026-06-19/12-46-33/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/12-46-33 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/12-46-33/.hydra/overrides.yaml b/outputs/2026-06-19/12-46-33/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9a9470fcb171b9de7af473034d3f72848dcf0e09 --- /dev/null +++ b/outputs/2026-06-19/12-46-33/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/12-46-33/main.log b/outputs/2026-06-19/12-46-33/main.log new file mode 100644 index 0000000000000000000000000000000000000000..d9e2f76eb6ea4c75dc25b4221fe64cc381d71d38 --- /dev/null +++ b/outputs/2026-06-19/12-46-33/main.log @@ -0,0 +1,5 @@ +[2026-06-19 12:47:18,207][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-19 12:48:13,606][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 12:48:17,646][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 12:46:33,052][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-22 12:46:33,491][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/12-46-34/.hydra/config.yaml b/outputs/2026-06-19/12-46-34/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ac3697b92d3aaa224d8690827ce376ca37c756e0 --- /dev/null +++ b/outputs/2026-06-19/12-46-34/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 4.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/12-46-34/.hydra/hydra.yaml b/outputs/2026-06-19/12-46-34/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..628ca91751af2dabd9cc2aa500eeca56657ab30a --- /dev/null +++ b/outputs/2026-06-19/12-46-34/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/12-46-34 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/12-46-34/.hydra/overrides.yaml b/outputs/2026-06-19/12-46-34/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9a9470fcb171b9de7af473034d3f72848dcf0e09 --- /dev/null +++ b/outputs/2026-06-19/12-46-34/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ditlr4e5_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=4.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/12-46-34/main.log b/outputs/2026-06-19/12-46-34/main.log new file mode 100644 index 0000000000000000000000000000000000000000..206b870550710ced25560a870f27080f40e8f2d3 --- /dev/null +++ b/outputs/2026-06-19/12-46-34/main.log @@ -0,0 +1,5 @@ +[2026-06-19 12:47:21,694][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-19 12:48:13,571][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 12:48:17,646][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 12:46:33,051][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-22 12:46:34,530][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/12-46-48/.hydra/config.yaml b/outputs/2026-06-19/12-46-48/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5f1434a7d8c57452d4e5434eff3f576e8af06352 --- /dev/null +++ b/outputs/2026-06-19/12-46-48/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 0 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/12-46-48/.hydra/hydra.yaml b/outputs/2026-06-19/12-46-48/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..173a387de8dd533903fbdebf1c3330d2080992fd --- /dev/null +++ b/outputs/2026-06-19/12-46-48/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=0 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=0,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/12-46-48 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/12-46-48/.hydra/overrides.yaml b/outputs/2026-06-19/12-46-48/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cd100937f0dd08e21d35eafd3b5f1531a2ba3251 --- /dev/null +++ b/outputs/2026-06-19/12-46-48/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=0 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/12-46-48/main.log b/outputs/2026-06-19/12-46-48/main.log new file mode 100644 index 0000000000000000000000000000000000000000..4988beea5b0a23ad332dac9901c6ce26a5cbd7a9 --- /dev/null +++ b/outputs/2026-06-19/12-46-48/main.log @@ -0,0 +1,6 @@ +[2026-06-19 12:47:33,363][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-19 12:47:33,367][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-19 12:48:17,981][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 12:48:18,057][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 12:48:21,686][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 12:48:21,686][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-19/12-46-49/.hydra/config.yaml b/outputs/2026-06-19/12-46-49/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5f1434a7d8c57452d4e5434eff3f576e8af06352 --- /dev/null +++ b/outputs/2026-06-19/12-46-49/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 0 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/12-46-49/.hydra/hydra.yaml b/outputs/2026-06-19/12-46-49/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0c7c669055405773602805740d14b6eb0a58173b --- /dev/null +++ b/outputs/2026-06-19/12-46-49/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=0 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=0,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/12-46-49 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/12-46-49/.hydra/overrides.yaml b/outputs/2026-06-19/12-46-49/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cd100937f0dd08e21d35eafd3b5f1531a2ba3251 --- /dev/null +++ b/outputs/2026-06-19/12-46-49/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=0 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/12-46-49/main.log b/outputs/2026-06-19/12-46-49/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e0916454afe22a44a4f7fd997be2bae2da0d040b --- /dev/null +++ b/outputs/2026-06-19/12-46-49/main.log @@ -0,0 +1,6 @@ +[2026-06-19 12:47:33,356][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-19 12:47:33,382][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-19 12:48:18,049][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 12:48:18,091][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 12:48:21,686][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 12:48:21,686][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-19/12-46-50/.hydra/config.yaml b/outputs/2026-06-19/12-46-50/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5f1434a7d8c57452d4e5434eff3f576e8af06352 --- /dev/null +++ b/outputs/2026-06-19/12-46-50/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 0 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/12-46-50/.hydra/hydra.yaml b/outputs/2026-06-19/12-46-50/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..33c11df5201b886565adae260f759bf0c3bfaaaa --- /dev/null +++ b/outputs/2026-06-19/12-46-50/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=0 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=0,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/12-46-50 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/12-46-50/.hydra/overrides.yaml b/outputs/2026-06-19/12-46-50/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cd100937f0dd08e21d35eafd3b5f1531a2ba3251 --- /dev/null +++ b/outputs/2026-06-19/12-46-50/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=0 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/12-46-50/main.log b/outputs/2026-06-19/12-46-50/main.log new file mode 100644 index 0000000000000000000000000000000000000000..24bb205acc262eff0a2d112e17b92eb70fef3a2b --- /dev/null +++ b/outputs/2026-06-19/12-46-50/main.log @@ -0,0 +1,3 @@ +[2026-06-19 12:47:34,003][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-19 12:48:16,940][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 12:48:21,686][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-19/12-46-51/.hydra/config.yaml b/outputs/2026-06-19/12-46-51/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5f1434a7d8c57452d4e5434eff3f576e8af06352 --- /dev/null +++ b/outputs/2026-06-19/12-46-51/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 0 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/12-46-51/.hydra/hydra.yaml b/outputs/2026-06-19/12-46-51/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..935b967d68e2f22601c22b5382ae35993cd04553 --- /dev/null +++ b/outputs/2026-06-19/12-46-51/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=0 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=0,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/12-46-51 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/12-46-51/.hydra/overrides.yaml b/outputs/2026-06-19/12-46-51/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cd100937f0dd08e21d35eafd3b5f1531a2ba3251 --- /dev/null +++ b/outputs/2026-06-19/12-46-51/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=0 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/12-46-51/main.log b/outputs/2026-06-19/12-46-51/main.log new file mode 100644 index 0000000000000000000000000000000000000000..b9e3901b795c1ce3bdb9ef18780c3def2e3e4d57 --- /dev/null +++ b/outputs/2026-06-19/12-46-51/main.log @@ -0,0 +1,3 @@ +[2026-06-19 12:47:33,379][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-19 12:48:18,067][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 12:48:21,686][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-19/12-46-52/.hydra/config.yaml b/outputs/2026-06-19/12-46-52/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5f1434a7d8c57452d4e5434eff3f576e8af06352 --- /dev/null +++ b/outputs/2026-06-19/12-46-52/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 0 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/12-46-52/.hydra/hydra.yaml b/outputs/2026-06-19/12-46-52/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..892ecd12491b6ead11e8f9fab969a46ae7259763 --- /dev/null +++ b/outputs/2026-06-19/12-46-52/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=0 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=0,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/12-46-52 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/12-46-52/.hydra/overrides.yaml b/outputs/2026-06-19/12-46-52/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cd100937f0dd08e21d35eafd3b5f1531a2ba3251 --- /dev/null +++ b/outputs/2026-06-19/12-46-52/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=0 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/12-46-52/main.log b/outputs/2026-06-19/12-46-52/main.log new file mode 100644 index 0000000000000000000000000000000000000000..fb557cacc9aa042b65336685beee6372c30b1bb2 --- /dev/null +++ b/outputs/2026-06-19/12-46-52/main.log @@ -0,0 +1,3 @@ +[2026-06-19 12:47:33,385][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-19 12:48:16,960][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 12:48:21,686][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-19/12-46-53/.hydra/config.yaml b/outputs/2026-06-19/12-46-53/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5f1434a7d8c57452d4e5434eff3f576e8af06352 --- /dev/null +++ b/outputs/2026-06-19/12-46-53/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 0 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/12-46-53/.hydra/hydra.yaml b/outputs/2026-06-19/12-46-53/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f0d93acaaf9aae51aa27843e882db495f0ae1903 --- /dev/null +++ b/outputs/2026-06-19/12-46-53/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=0 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=0,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/12-46-53 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/12-46-53/.hydra/overrides.yaml b/outputs/2026-06-19/12-46-53/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cd100937f0dd08e21d35eafd3b5f1531a2ba3251 --- /dev/null +++ b/outputs/2026-06-19/12-46-53/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=0 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/12-46-53/main.log b/outputs/2026-06-19/12-46-53/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e39079c6bb55ce4a8483b0c716a0abf685c1c984 --- /dev/null +++ b/outputs/2026-06-19/12-46-53/main.log @@ -0,0 +1,28 @@ +[2026-06-19 12:47:30,487][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-19 12:47:30,651][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-19 12:47:30,651][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-19 12:47:30,652][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-19 12:47:30,652][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-19 12:47:30,652][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-19 12:47:33,380][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-19 12:47:39,783][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-19 12:47:55,031][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/checkpoints/last.ckpt +[2026-06-19 12:48:17,958][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 12:48:19,575][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +1.0 B Trainable params +231 M Non-trainable params +1.3 B Total params +5,107.144 Total estimated model params size (MB) +[2026-06-19 12:48:20,760][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_anchor_d4r2_route_ntok8_8a100_bs8_global64_350k/checkpoints/last.ckpt +[2026-06-19 12:48:21,686][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-21 19:35:52,927][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=350000` reached. diff --git a/outputs/2026-06-19/13-50-53/.hydra/config.yaml b/outputs/2026-06-19/13-50-53/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/13-50-53/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/13-50-53/.hydra/hydra.yaml b/outputs/2026-06-19/13-50-53/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3e9b3acf6cf0f5e97994e4ad49571a0ec8debbf8 --- /dev/null +++ b/outputs/2026-06-19/13-50-53/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/13-50-53 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/13-50-53/.hydra/overrides.yaml b/outputs/2026-06-19/13-50-53/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/13-50-53/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/13-50-53/main.log b/outputs/2026-06-19/13-50-53/main.log new file mode 100644 index 0000000000000000000000000000000000000000..38d3566f3dce1c0e1ec87d763e1d21adf6149b5b --- /dev/null +++ b/outputs/2026-06-19/13-50-53/main.log @@ -0,0 +1 @@ +[2026-06-19 13:51:44,736][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 diff --git a/outputs/2026-06-19/13-50-54/.hydra/config.yaml b/outputs/2026-06-19/13-50-54/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/13-50-54/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/13-50-54/.hydra/hydra.yaml b/outputs/2026-06-19/13-50-54/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4808628f916552b8b85d6b47b395b30469a90810 --- /dev/null +++ b/outputs/2026-06-19/13-50-54/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/13-50-54 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/13-50-54/.hydra/overrides.yaml b/outputs/2026-06-19/13-50-54/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/13-50-54/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/13-50-54/main.log b/outputs/2026-06-19/13-50-54/main.log new file mode 100644 index 0000000000000000000000000000000000000000..d11f414f6dccf5b0f5caefcfe9b24edda43ffcd1 --- /dev/null +++ b/outputs/2026-06-19/13-50-54/main.log @@ -0,0 +1 @@ +[2026-06-19 13:51:44,749][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 diff --git a/outputs/2026-06-19/13-50-55/.hydra/config.yaml b/outputs/2026-06-19/13-50-55/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/13-50-55/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/13-50-55/.hydra/hydra.yaml b/outputs/2026-06-19/13-50-55/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a5aaaad5e172c4c5ee69fa79c43ffb3b3e6060e1 --- /dev/null +++ b/outputs/2026-06-19/13-50-55/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/13-50-55 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/13-50-55/.hydra/overrides.yaml b/outputs/2026-06-19/13-50-55/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/13-50-55/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/13-50-55/main.log b/outputs/2026-06-19/13-50-55/main.log new file mode 100644 index 0000000000000000000000000000000000000000..8979a4bf03e3a87518934b159a2ad5fa9ca31159 --- /dev/null +++ b/outputs/2026-06-19/13-50-55/main.log @@ -0,0 +1 @@ +[2026-06-19 13:51:44,742][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 diff --git a/outputs/2026-06-19/13-50-56/.hydra/config.yaml b/outputs/2026-06-19/13-50-56/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/13-50-56/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/13-50-56/.hydra/hydra.yaml b/outputs/2026-06-19/13-50-56/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f5a79ca50eb0f561b96a1eb1897a0bd390acc8db --- /dev/null +++ b/outputs/2026-06-19/13-50-56/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/13-50-56 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/13-50-56/.hydra/overrides.yaml b/outputs/2026-06-19/13-50-56/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/13-50-56/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/13-50-56/main.log b/outputs/2026-06-19/13-50-56/main.log new file mode 100644 index 0000000000000000000000000000000000000000..a7c11afd2b4a1e364b1a1fad020b700a866c977e --- /dev/null +++ b/outputs/2026-06-19/13-50-56/main.log @@ -0,0 +1 @@ +[2026-06-19 13:51:44,744][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 diff --git a/outputs/2026-06-19/13-50-57/.hydra/config.yaml b/outputs/2026-06-19/13-50-57/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/13-50-57/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/13-50-57/.hydra/hydra.yaml b/outputs/2026-06-19/13-50-57/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cbe47ddf398e3578e162feee375e1b84b6c4766f --- /dev/null +++ b/outputs/2026-06-19/13-50-57/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/13-50-57 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/13-50-57/.hydra/overrides.yaml b/outputs/2026-06-19/13-50-57/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/13-50-57/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/13-50-57/main.log b/outputs/2026-06-19/13-50-57/main.log new file mode 100644 index 0000000000000000000000000000000000000000..fa72a7cec08fe579041828edefb8c55b0006adb6 --- /dev/null +++ b/outputs/2026-06-19/13-50-57/main.log @@ -0,0 +1 @@ +[2026-06-19 13:51:44,737][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 diff --git a/outputs/2026-06-19/13-50-58/.hydra/config.yaml b/outputs/2026-06-19/13-50-58/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/13-50-58/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/13-50-58/.hydra/hydra.yaml b/outputs/2026-06-19/13-50-58/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3f47b52b95efc2804bcfc0cfd45c514a620fa588 --- /dev/null +++ b/outputs/2026-06-19/13-50-58/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/13-50-58 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/13-50-58/.hydra/overrides.yaml b/outputs/2026-06-19/13-50-58/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/13-50-58/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/13-50-58/main.log b/outputs/2026-06-19/13-50-58/main.log new file mode 100644 index 0000000000000000000000000000000000000000..ce4c6f23e9ec2ae8a0f9ca77ab277b02010d567b --- /dev/null +++ b/outputs/2026-06-19/13-50-58/main.log @@ -0,0 +1 @@ +[2026-06-19 13:51:44,765][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 diff --git a/outputs/2026-06-19/13-50-59/.hydra/config.yaml b/outputs/2026-06-19/13-50-59/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/13-50-59/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/13-50-59/.hydra/hydra.yaml b/outputs/2026-06-19/13-50-59/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b15331e950014b037dcfd17fc2915965b24d6896 --- /dev/null +++ b/outputs/2026-06-19/13-50-59/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/13-50-59 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/13-50-59/.hydra/overrides.yaml b/outputs/2026-06-19/13-50-59/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/13-50-59/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/13-50-59/main.log b/outputs/2026-06-19/13-50-59/main.log new file mode 100644 index 0000000000000000000000000000000000000000..1616ed9db0ebb14aaea7332a8ebe259a245530ee --- /dev/null +++ b/outputs/2026-06-19/13-50-59/main.log @@ -0,0 +1 @@ +[2026-06-19 13:51:44,744][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 diff --git a/outputs/2026-06-19/13-51-00/.hydra/config.yaml b/outputs/2026-06-19/13-51-00/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/13-51-00/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/13-51-00/.hydra/hydra.yaml b/outputs/2026-06-19/13-51-00/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..58e297d8d6e78f6c2ecdccf0c290bba928465b1d --- /dev/null +++ b/outputs/2026-06-19/13-51-00/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/13-51-00 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/13-51-00/.hydra/overrides.yaml b/outputs/2026-06-19/13-51-00/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/13-51-00/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/13-51-00/main.log b/outputs/2026-06-19/13-51-00/main.log new file mode 100644 index 0000000000000000000000000000000000000000..c98d6fbd653d133639c0305db3dd39616efeccf0 --- /dev/null +++ b/outputs/2026-06-19/13-51-00/main.log @@ -0,0 +1,14 @@ +[2026-06-19 13:51:38,447][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-19 13:51:38,588][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-19 13:51:38,589][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-19 13:51:38,589][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-19 13:51:38,589][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-19 13:51:38,589][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-19 13:51:41,316][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-19 13:51:42,540][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-19 13:51:44,753][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-19 13:51:50,732][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + diff --git a/outputs/2026-06-19/14-11-20/.hydra/config.yaml b/outputs/2026-06-19/14-11-20/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..00a4819247bb946a4bde3e323b2ace07435b877f --- /dev/null +++ b/outputs/2026-06-19/14-11-20/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/14-11-20/.hydra/hydra.yaml b/outputs/2026-06-19/14-11-20/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7e1f5c54cf62d620f28d96e7e3cad6e5e360a34d --- /dev/null +++ b/outputs/2026-06-19/14-11-20/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/14-11-20 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/14-11-20/.hydra/overrides.yaml b/outputs/2026-06-19/14-11-20/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c327e3e8696cb770e644c8cdb4761d5660506390 --- /dev/null +++ b/outputs/2026-06-19/14-11-20/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/14-11-20/main.log b/outputs/2026-06-19/14-11-20/main.log new file mode 100644 index 0000000000000000000000000000000000000000..03b4bb1c191acfef8452948d96f11d9d785a1492 --- /dev/null +++ b/outputs/2026-06-19/14-11-20/main.log @@ -0,0 +1,3 @@ +[2026-06-19 14:12:05,934][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-19 14:12:53,020][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 14:12:58,367][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-19/14-11-21/.hydra/config.yaml b/outputs/2026-06-19/14-11-21/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..00a4819247bb946a4bde3e323b2ace07435b877f --- /dev/null +++ b/outputs/2026-06-19/14-11-21/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/14-11-21/.hydra/hydra.yaml b/outputs/2026-06-19/14-11-21/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b777e83bee2e3b9e6868f71b6d2f20e85d8c8c9b --- /dev/null +++ b/outputs/2026-06-19/14-11-21/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/14-11-21 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/14-11-21/.hydra/overrides.yaml b/outputs/2026-06-19/14-11-21/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c327e3e8696cb770e644c8cdb4761d5660506390 --- /dev/null +++ b/outputs/2026-06-19/14-11-21/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/14-11-21/main.log b/outputs/2026-06-19/14-11-21/main.log new file mode 100644 index 0000000000000000000000000000000000000000..ff10450f635d1c0817ce591bf417b56d8e76512c --- /dev/null +++ b/outputs/2026-06-19/14-11-21/main.log @@ -0,0 +1,6 @@ +[2026-06-19 14:12:05,480][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-19 14:12:05,483][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-19 14:12:52,557][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 14:12:53,032][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 14:12:58,367][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 14:12:58,367][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-19/14-11-22/.hydra/config.yaml b/outputs/2026-06-19/14-11-22/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..00a4819247bb946a4bde3e323b2ace07435b877f --- /dev/null +++ b/outputs/2026-06-19/14-11-22/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/14-11-22/.hydra/hydra.yaml b/outputs/2026-06-19/14-11-22/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..778a8459e32baae885eae0484a25caee667eeb43 --- /dev/null +++ b/outputs/2026-06-19/14-11-22/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/14-11-22 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/14-11-22/.hydra/overrides.yaml b/outputs/2026-06-19/14-11-22/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c327e3e8696cb770e644c8cdb4761d5660506390 --- /dev/null +++ b/outputs/2026-06-19/14-11-22/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/14-11-22/main.log b/outputs/2026-06-19/14-11-22/main.log new file mode 100644 index 0000000000000000000000000000000000000000..03b3d50ce6910bb0365c8bc54c8c9fe1d02496e9 --- /dev/null +++ b/outputs/2026-06-19/14-11-22/main.log @@ -0,0 +1,3 @@ +[2026-06-19 14:12:05,578][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-19 14:12:53,039][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 14:12:58,367][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-19/14-11-23/.hydra/config.yaml b/outputs/2026-06-19/14-11-23/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..00a4819247bb946a4bde3e323b2ace07435b877f --- /dev/null +++ b/outputs/2026-06-19/14-11-23/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/14-11-23/.hydra/hydra.yaml b/outputs/2026-06-19/14-11-23/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..42d02e8c640e40854cda299154f9da25ac0f211f --- /dev/null +++ b/outputs/2026-06-19/14-11-23/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/14-11-23 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/14-11-23/.hydra/overrides.yaml b/outputs/2026-06-19/14-11-23/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c327e3e8696cb770e644c8cdb4761d5660506390 --- /dev/null +++ b/outputs/2026-06-19/14-11-23/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/14-11-23/main.log b/outputs/2026-06-19/14-11-23/main.log new file mode 100644 index 0000000000000000000000000000000000000000..a25b3dd59f6b0b5ba76592a144d0b1f1ac6fedc4 --- /dev/null +++ b/outputs/2026-06-19/14-11-23/main.log @@ -0,0 +1,3 @@ +[2026-06-19 14:12:05,473][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-19 14:12:53,039][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 14:12:58,367][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-19/14-11-24/.hydra/config.yaml b/outputs/2026-06-19/14-11-24/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..00a4819247bb946a4bde3e323b2ace07435b877f --- /dev/null +++ b/outputs/2026-06-19/14-11-24/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/14-11-24/.hydra/hydra.yaml b/outputs/2026-06-19/14-11-24/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5ae5d18eb32951e955a315195809838b07929e0c --- /dev/null +++ b/outputs/2026-06-19/14-11-24/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/14-11-24 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/14-11-24/.hydra/overrides.yaml b/outputs/2026-06-19/14-11-24/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c327e3e8696cb770e644c8cdb4761d5660506390 --- /dev/null +++ b/outputs/2026-06-19/14-11-24/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/14-11-24/main.log b/outputs/2026-06-19/14-11-24/main.log new file mode 100644 index 0000000000000000000000000000000000000000..6a1ea010b2c70c74c69403671941a6dc16fdd3ca --- /dev/null +++ b/outputs/2026-06-19/14-11-24/main.log @@ -0,0 +1,28 @@ +[2026-06-19 14:12:02,642][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-19 14:12:02,779][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-19 14:12:02,779][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-19 14:12:02,779][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-19 14:12:02,779][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-19 14:12:02,780][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-19 14:12:05,492][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-19 14:12:11,974][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-19 14:12:27,493][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/checkpoints/last.ckpt +[2026-06-19 14:12:52,555][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 14:12:56,384][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +1.0 B Trainable params +231 M Non-trainable params +1.3 B Total params +5,107.144 Total estimated model params size (MB) +[2026-06-19 14:12:57,392][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/checkpoints/last.ckpt +[2026-06-19 14:12:58,367][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-21 12:57:55,245][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=350000` reached. diff --git a/outputs/2026-06-19/14-11-25/.hydra/config.yaml b/outputs/2026-06-19/14-11-25/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..00a4819247bb946a4bde3e323b2ace07435b877f --- /dev/null +++ b/outputs/2026-06-19/14-11-25/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/14-11-25/.hydra/hydra.yaml b/outputs/2026-06-19/14-11-25/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..08379165b429c84666d34bd56c13e30ccd818074 --- /dev/null +++ b/outputs/2026-06-19/14-11-25/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/14-11-25 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/14-11-25/.hydra/overrides.yaml b/outputs/2026-06-19/14-11-25/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c327e3e8696cb770e644c8cdb4761d5660506390 --- /dev/null +++ b/outputs/2026-06-19/14-11-25/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/14-11-25/main.log b/outputs/2026-06-19/14-11-25/main.log new file mode 100644 index 0000000000000000000000000000000000000000..a0f8fffe8ba537d974437c806d4937e3fa26c35f --- /dev/null +++ b/outputs/2026-06-19/14-11-25/main.log @@ -0,0 +1,3 @@ +[2026-06-19 14:12:05,488][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-19 14:12:53,037][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 14:12:58,367][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-19/14-11-26/.hydra/config.yaml b/outputs/2026-06-19/14-11-26/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..00a4819247bb946a4bde3e323b2ace07435b877f --- /dev/null +++ b/outputs/2026-06-19/14-11-26/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/14-11-26/.hydra/hydra.yaml b/outputs/2026-06-19/14-11-26/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..39cdcb95392f79499a6493385a86996af1ea9fcd --- /dev/null +++ b/outputs/2026-06-19/14-11-26/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/14-11-26 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/14-11-26/.hydra/overrides.yaml b/outputs/2026-06-19/14-11-26/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c327e3e8696cb770e644c8cdb4761d5660506390 --- /dev/null +++ b/outputs/2026-06-19/14-11-26/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_dynamic_a2r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/14-11-26/main.log b/outputs/2026-06-19/14-11-26/main.log new file mode 100644 index 0000000000000000000000000000000000000000..d2102f455c136cc9e9e6e0cd69bf2e0b3a8e3f2c --- /dev/null +++ b/outputs/2026-06-19/14-11-26/main.log @@ -0,0 +1,3 @@ +[2026-06-19 14:12:05,455][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-19 14:12:53,027][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 14:12:58,367][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-19/15-18-30/.hydra/config.yaml b/outputs/2026-06-19/15-18-30/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/15-18-30/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/15-18-30/.hydra/hydra.yaml b/outputs/2026-06-19/15-18-30/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a716fff2693ad1ea5c8c261e929782881ab5f49b --- /dev/null +++ b/outputs/2026-06-19/15-18-30/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/15-18-30 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/15-18-30/.hydra/overrides.yaml b/outputs/2026-06-19/15-18-30/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/15-18-30/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/15-18-30/main.log b/outputs/2026-06-19/15-18-30/main.log new file mode 100644 index 0000000000000000000000000000000000000000..b3f8e49514325549894b25df199a624c1cbe6e9a --- /dev/null +++ b/outputs/2026-06-19/15-18-30/main.log @@ -0,0 +1,16 @@ +[2026-06-19 15:19:03,194][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-19 15:19:03,350][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-19 15:19:03,350][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-19 15:19:03,350][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-19 15:19:03,350][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-19 15:19:03,351][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-19 15:19:03,467][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-19 15:19:03,590][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-19 15:19:05,102][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-19 15:19:05,208][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-19 15:19:05,668][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-19 15:19:11,405][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + diff --git a/outputs/2026-06-19/15-18-31/.hydra/config.yaml b/outputs/2026-06-19/15-18-31/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/15-18-31/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/15-18-31/.hydra/hydra.yaml b/outputs/2026-06-19/15-18-31/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..856e66b5c32a8b969114891a46b6282ac21f286c --- /dev/null +++ b/outputs/2026-06-19/15-18-31/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/15-18-31 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/15-18-31/.hydra/overrides.yaml b/outputs/2026-06-19/15-18-31/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/15-18-31/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/15-18-31/main.log b/outputs/2026-06-19/15-18-31/main.log new file mode 100644 index 0000000000000000000000000000000000000000..3405f7f5639dcfc2b35029b5f69affbb062aede9 --- /dev/null +++ b/outputs/2026-06-19/15-18-31/main.log @@ -0,0 +1,2 @@ +[2026-06-19 15:19:05,465][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-19 15:19:05,952][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 diff --git a/outputs/2026-06-19/15-18-33/.hydra/config.yaml b/outputs/2026-06-19/15-18-33/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/15-18-33/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/15-18-33/.hydra/hydra.yaml b/outputs/2026-06-19/15-18-33/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..53bc79e1c65849d08f14ac8d62a3ade72588874b --- /dev/null +++ b/outputs/2026-06-19/15-18-33/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/15-18-33 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/15-18-33/.hydra/overrides.yaml b/outputs/2026-06-19/15-18-33/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/15-18-33/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/15-18-33/main.log b/outputs/2026-06-19/15-18-33/main.log new file mode 100644 index 0000000000000000000000000000000000000000..53b5bea23d32cd8ea53699461ef468a73091884f --- /dev/null +++ b/outputs/2026-06-19/15-18-33/main.log @@ -0,0 +1 @@ +[2026-06-19 15:19:05,166][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 diff --git a/outputs/2026-06-19/15-18-34/.hydra/config.yaml b/outputs/2026-06-19/15-18-34/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/15-18-34/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/15-18-34/.hydra/hydra.yaml b/outputs/2026-06-19/15-18-34/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9899b6658c221bee3e0e49bc0413dbb0e2780102 --- /dev/null +++ b/outputs/2026-06-19/15-18-34/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/15-18-34 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/15-18-34/.hydra/overrides.yaml b/outputs/2026-06-19/15-18-34/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/15-18-34/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/15-18-34/main.log b/outputs/2026-06-19/15-18-34/main.log new file mode 100644 index 0000000000000000000000000000000000000000..8589897da675e2392db7d728f2a6c96c8fe44949 --- /dev/null +++ b/outputs/2026-06-19/15-18-34/main.log @@ -0,0 +1 @@ +[2026-06-19 15:19:05,048][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 diff --git a/outputs/2026-06-19/15-18-35/.hydra/config.yaml b/outputs/2026-06-19/15-18-35/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/15-18-35/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/15-18-35/.hydra/hydra.yaml b/outputs/2026-06-19/15-18-35/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..194ecf2cb2a885eb796a679560ed5a1544bf3c51 --- /dev/null +++ b/outputs/2026-06-19/15-18-35/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/15-18-35 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/15-18-35/.hydra/overrides.yaml b/outputs/2026-06-19/15-18-35/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/15-18-35/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/15-18-35/main.log b/outputs/2026-06-19/15-18-35/main.log new file mode 100644 index 0000000000000000000000000000000000000000..45cf754d1e6cf633251893aa63e3a47a1a557cc2 --- /dev/null +++ b/outputs/2026-06-19/15-18-35/main.log @@ -0,0 +1 @@ +[2026-06-19 15:19:06,637][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 diff --git a/outputs/2026-06-19/15-19-45/.hydra/config.yaml b/outputs/2026-06-19/15-19-45/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c59a632cf78c535166e69ca7f9d7cc18ae1e3f3b --- /dev/null +++ b/outputs/2026-06-19/15-19-45/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/15-19-45/.hydra/hydra.yaml b/outputs/2026-06-19/15-19-45/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..be048b5b4af5419ae5dd62de51a210573bcfa36d --- /dev/null +++ b/outputs/2026-06-19/15-19-45/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/15-19-45 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/15-19-45/.hydra/overrides.yaml b/outputs/2026-06-19/15-19-45/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7dc9321e7110bcf1e972bf4ff4e64f393e83fa21 --- /dev/null +++ b/outputs/2026-06-19/15-19-45/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/15-19-45/main.log b/outputs/2026-06-19/15-19-45/main.log new file mode 100644 index 0000000000000000000000000000000000000000..5e2c710d874aee5e2025e2e1b8d5cbc46589e116 --- /dev/null +++ b/outputs/2026-06-19/15-19-45/main.log @@ -0,0 +1,17 @@ +[2026-06-19 15:20:16,054][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-19 15:20:16,200][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-19 15:20:16,200][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-19 15:20:16,200][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-19 15:20:16,201][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-19 15:20:16,201][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-19 15:20:16,302][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-19 15:20:16,415][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-19 15:20:18,103][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-19 15:20:18,105][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-19 15:20:18,427][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-19 15:20:18,606][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-19 15:20:23,766][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + diff --git a/outputs/2026-06-19/15-19-46/.hydra/config.yaml b/outputs/2026-06-19/15-19-46/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c59a632cf78c535166e69ca7f9d7cc18ae1e3f3b --- /dev/null +++ b/outputs/2026-06-19/15-19-46/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/15-19-46/.hydra/hydra.yaml b/outputs/2026-06-19/15-19-46/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fe81399a285d8cc34b86514c6583d811b7216129 --- /dev/null +++ b/outputs/2026-06-19/15-19-46/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/15-19-46 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/15-19-46/.hydra/overrides.yaml b/outputs/2026-06-19/15-19-46/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7dc9321e7110bcf1e972bf4ff4e64f393e83fa21 --- /dev/null +++ b/outputs/2026-06-19/15-19-46/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/15-19-46/main.log b/outputs/2026-06-19/15-19-46/main.log new file mode 100644 index 0000000000000000000000000000000000000000..71ffbfb6275df57a213495ee837e2ebaaa49bd9c --- /dev/null +++ b/outputs/2026-06-19/15-19-46/main.log @@ -0,0 +1,2 @@ +[2026-06-19 15:20:18,219][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-19 15:20:19,496][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 diff --git a/outputs/2026-06-19/15-19-47/.hydra/config.yaml b/outputs/2026-06-19/15-19-47/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c59a632cf78c535166e69ca7f9d7cc18ae1e3f3b --- /dev/null +++ b/outputs/2026-06-19/15-19-47/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/15-19-47/.hydra/hydra.yaml b/outputs/2026-06-19/15-19-47/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f4ba125198c3f7deea52e33ed1a7debe9b93c409 --- /dev/null +++ b/outputs/2026-06-19/15-19-47/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/15-19-47 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/15-19-47/.hydra/overrides.yaml b/outputs/2026-06-19/15-19-47/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7dc9321e7110bcf1e972bf4ff4e64f393e83fa21 --- /dev/null +++ b/outputs/2026-06-19/15-19-47/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/15-19-47/main.log b/outputs/2026-06-19/15-19-47/main.log new file mode 100644 index 0000000000000000000000000000000000000000..64b4ecda91d3b00f121b54fd626a2b5f78fb414a --- /dev/null +++ b/outputs/2026-06-19/15-19-47/main.log @@ -0,0 +1 @@ +[2026-06-19 15:20:18,869][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 diff --git a/outputs/2026-06-19/15-19-48/.hydra/config.yaml b/outputs/2026-06-19/15-19-48/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c59a632cf78c535166e69ca7f9d7cc18ae1e3f3b --- /dev/null +++ b/outputs/2026-06-19/15-19-48/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/15-19-48/.hydra/hydra.yaml b/outputs/2026-06-19/15-19-48/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b551fa5f59fdd4568a97763cd890429afdc27c59 --- /dev/null +++ b/outputs/2026-06-19/15-19-48/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/15-19-48 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/15-19-48/.hydra/overrides.yaml b/outputs/2026-06-19/15-19-48/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7dc9321e7110bcf1e972bf4ff4e64f393e83fa21 --- /dev/null +++ b/outputs/2026-06-19/15-19-48/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/15-19-48/main.log b/outputs/2026-06-19/15-19-48/main.log new file mode 100644 index 0000000000000000000000000000000000000000..c17189f630c17c341bf3a86b340472bd6870d018 --- /dev/null +++ b/outputs/2026-06-19/15-19-48/main.log @@ -0,0 +1 @@ +[2026-06-19 15:20:18,197][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 diff --git a/outputs/2026-06-19/15-54-21/.hydra/config.yaml b/outputs/2026-06-19/15-54-21/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9869f51c161adbbd6d3c059d1e67ed8908232dc6 --- /dev/null +++ b/outputs/2026-06-19/15-54-21/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/15-54-21/.hydra/hydra.yaml b/outputs/2026-06-19/15-54-21/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fa81110577f7cd47e7a41cb3dba4c5887d64175b --- /dev/null +++ b/outputs/2026-06-19/15-54-21/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/15-54-21 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/15-54-21/.hydra/overrides.yaml b/outputs/2026-06-19/15-54-21/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5e998fce7e79fdbf9bdb2057d2f99d7858591ce3 --- /dev/null +++ b/outputs/2026-06-19/15-54-21/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/15-54-21/main.log b/outputs/2026-06-19/15-54-21/main.log new file mode 100644 index 0000000000000000000000000000000000000000..9917a4a2e3ed60da32dd0a7713fdad8cf873c62f --- /dev/null +++ b/outputs/2026-06-19/15-54-21/main.log @@ -0,0 +1,15 @@ +[2026-06-19 15:55:07,715][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-19 15:55:07,746][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-19 15:55:11,454][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-19 15:55:55,569][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 15:55:55,608][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 15:55:55,645][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 15:56:00,119][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 15:56:00,120][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 15:56:00,120][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 15:54:35,174][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-22 15:54:35,175][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-22 15:54:35,174][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-22 15:54:35,175][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-22 15:54:35,173][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-22 15:54:35,176][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/15-54-22/.hydra/config.yaml b/outputs/2026-06-19/15-54-22/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9869f51c161adbbd6d3c059d1e67ed8908232dc6 --- /dev/null +++ b/outputs/2026-06-19/15-54-22/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/15-54-22/.hydra/hydra.yaml b/outputs/2026-06-19/15-54-22/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..82f4764e07490c8f867af37fd8a518e97fcf16ac --- /dev/null +++ b/outputs/2026-06-19/15-54-22/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/15-54-22 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/15-54-22/.hydra/overrides.yaml b/outputs/2026-06-19/15-54-22/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5e998fce7e79fdbf9bdb2057d2f99d7858591ce3 --- /dev/null +++ b/outputs/2026-06-19/15-54-22/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/15-54-22/main.log b/outputs/2026-06-19/15-54-22/main.log new file mode 100644 index 0000000000000000000000000000000000000000..c76ae0bb36dae1a643d1b21aa45918f0996e13ce --- /dev/null +++ b/outputs/2026-06-19/15-54-22/main.log @@ -0,0 +1,5 @@ +[2026-06-19 15:55:08,475][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-19 15:55:55,589][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 15:56:00,120][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 15:54:35,175][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-22 15:54:35,176][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/15-54-23/.hydra/config.yaml b/outputs/2026-06-19/15-54-23/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9869f51c161adbbd6d3c059d1e67ed8908232dc6 --- /dev/null +++ b/outputs/2026-06-19/15-54-23/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/15-54-23/.hydra/hydra.yaml b/outputs/2026-06-19/15-54-23/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..26dff8841a5c25597a805199d2fd214f0dafcd14 --- /dev/null +++ b/outputs/2026-06-19/15-54-23/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/15-54-23 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/15-54-23/.hydra/overrides.yaml b/outputs/2026-06-19/15-54-23/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5e998fce7e79fdbf9bdb2057d2f99d7858591ce3 --- /dev/null +++ b/outputs/2026-06-19/15-54-23/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/15-54-23/main.log b/outputs/2026-06-19/15-54-23/main.log new file mode 100644 index 0000000000000000000000000000000000000000..0cc5a3c70a5b469c6ca9187a33a6835b68b668cd --- /dev/null +++ b/outputs/2026-06-19/15-54-23/main.log @@ -0,0 +1,5 @@ +[2026-06-19 15:55:07,944][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-19 15:55:55,645][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 15:56:00,119][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 15:54:35,174][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-22 15:54:35,175][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/15-54-24/.hydra/config.yaml b/outputs/2026-06-19/15-54-24/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9869f51c161adbbd6d3c059d1e67ed8908232dc6 --- /dev/null +++ b/outputs/2026-06-19/15-54-24/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/15-54-24/.hydra/hydra.yaml b/outputs/2026-06-19/15-54-24/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..2b16130094e0fe6e8f315e029aa8175765f4b44b --- /dev/null +++ b/outputs/2026-06-19/15-54-24/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/15-54-24 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/15-54-24/.hydra/overrides.yaml b/outputs/2026-06-19/15-54-24/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5e998fce7e79fdbf9bdb2057d2f99d7858591ce3 --- /dev/null +++ b/outputs/2026-06-19/15-54-24/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/15-54-24/main.log b/outputs/2026-06-19/15-54-24/main.log new file mode 100644 index 0000000000000000000000000000000000000000..d2980ecf73db3809ae4704bf0707451cc31e65d5 --- /dev/null +++ b/outputs/2026-06-19/15-54-24/main.log @@ -0,0 +1,5 @@ +[2026-06-19 15:55:07,720][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-19 15:55:55,632][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 15:56:00,119][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 15:54:35,174][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-22 15:54:35,176][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/15-54-25/.hydra/config.yaml b/outputs/2026-06-19/15-54-25/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9869f51c161adbbd6d3c059d1e67ed8908232dc6 --- /dev/null +++ b/outputs/2026-06-19/15-54-25/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/15-54-25/.hydra/hydra.yaml b/outputs/2026-06-19/15-54-25/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5fdd079b000b6da3736ca12d6a0fe86504fbf947 --- /dev/null +++ b/outputs/2026-06-19/15-54-25/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/15-54-25 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/15-54-25/.hydra/overrides.yaml b/outputs/2026-06-19/15-54-25/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5e998fce7e79fdbf9bdb2057d2f99d7858591ce3 --- /dev/null +++ b/outputs/2026-06-19/15-54-25/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/15-54-25/main.log b/outputs/2026-06-19/15-54-25/main.log new file mode 100644 index 0000000000000000000000000000000000000000..39c626b22b2c9126f0f544535ff4b4ecaa394845 --- /dev/null +++ b/outputs/2026-06-19/15-54-25/main.log @@ -0,0 +1,5 @@ +[2026-06-19 15:55:07,727][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-19 15:55:55,129][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 15:56:00,120][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 15:54:35,174][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-22 15:54:35,176][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/15-54-26/.hydra/config.yaml b/outputs/2026-06-19/15-54-26/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9869f51c161adbbd6d3c059d1e67ed8908232dc6 --- /dev/null +++ b/outputs/2026-06-19/15-54-26/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/15-54-26/.hydra/hydra.yaml b/outputs/2026-06-19/15-54-26/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a4fe6a82310588eaaa21f6d606f0eeeb651aa402 --- /dev/null +++ b/outputs/2026-06-19/15-54-26/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/15-54-26 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/15-54-26/.hydra/overrides.yaml b/outputs/2026-06-19/15-54-26/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5e998fce7e79fdbf9bdb2057d2f99d7858591ce3 --- /dev/null +++ b/outputs/2026-06-19/15-54-26/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/15-54-26/main.log b/outputs/2026-06-19/15-54-26/main.log new file mode 100644 index 0000000000000000000000000000000000000000..5b69e4f8b61e484e27b35b9bccc7ba23adb83d03 --- /dev/null +++ b/outputs/2026-06-19/15-54-26/main.log @@ -0,0 +1,29 @@ +[2026-06-19 15:55:04,888][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-19 15:55:06,098][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-19 15:55:06,098][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-19 15:55:06,098][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-19 15:55:06,098][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-19 15:55:06,099][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-19 15:55:08,275][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-19 15:55:15,852][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-19 15:55:29,583][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/checkpoints/last.ckpt +[2026-06-19 15:55:55,114][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 15:55:57,124][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +1.0 B Trainable params +231 M Non-trainable params +1.3 B Total params +5,107.144 Total estimated model params size (MB) +[2026-06-19 15:55:58,057][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8a100_bs8_global64_350k/checkpoints/last.ckpt +[2026-06-19 15:56:00,120][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 15:54:35,174][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-22 15:54:35,176][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/18-56-25/.hydra/config.yaml b/outputs/2026-06-19/18-56-25/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b47e3332559c53ed999c275a717b07d1982311f8 --- /dev/null +++ b/outputs/2026-06-19/18-56-25/.hydra/config.yaml @@ -0,0 +1,231 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: + - 3 + - 7 + - 11 + - 15 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/18-56-25/.hydra/hydra.yaml b/outputs/2026-06-19/18-56-25/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..6319433939100c419c3ca27b905ebfa878fc72d1 --- /dev/null +++ b/outputs/2026-06-19/18-56-25/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - algorithm.memory_cross_attn_layers=[3,7,11,15] + - ++algorithm.dememwm.injection.memory_hidden_size=512 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.injection.memory_hidden_size=512,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.memory_cross_attn_layers=[3,7,11,15],algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/18-56-25 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/18-56-25/.hydra/overrides.yaml b/outputs/2026-06-19/18-56-25/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8ee3288e706e04827ca476ead0592fea9c288ffa --- /dev/null +++ b/outputs/2026-06-19/18-56-25/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- algorithm.memory_cross_attn_layers=[3,7,11,15] +- ++algorithm.dememwm.injection.memory_hidden_size=512 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/18-56-25/main.log b/outputs/2026-06-19/18-56-25/main.log new file mode 100644 index 0000000000000000000000000000000000000000..dbc487c7e1fde877599da4f92c945c87ea609e49 --- /dev/null +++ b/outputs/2026-06-19/18-56-25/main.log @@ -0,0 +1,20 @@ +[2026-06-19 18:57:06,921][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-19 18:57:07,170][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-19 18:57:07,278][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-19 18:57:07,358][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-19 18:57:46,640][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 18:57:47,229][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 18:57:47,231][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 18:57:47,287][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 18:57:52,222][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 18:57:52,222][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 18:57:52,222][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 18:57:52,222][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 19:04:55,613][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-19 19:04:55,614][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-19 19:04:55,614][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-19 19:04:55,614][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-19 19:04:55,614][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-19 19:04:55,614][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-19 19:04:55,614][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-19 19:04:55,615][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/18-56-26/.hydra/config.yaml b/outputs/2026-06-19/18-56-26/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b47e3332559c53ed999c275a717b07d1982311f8 --- /dev/null +++ b/outputs/2026-06-19/18-56-26/.hydra/config.yaml @@ -0,0 +1,231 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + customized_validation: true + wo_updown: false + angle_range: 180 + pos_range: 8 +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_memory_dit + memory_token_cross_attention: true + memory_cross_attn_layers: + - 3 + - 7 + - 11 + - 15 + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + training_stage: stage_1 + curriculum: + enabled: true + full_stage_start_step: 20000 + freeze_vae: true + dit_freeze: + enabled: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + spatial_kernel_size: 3 + anchor: + enabled: true + anchor_indices: + - 0 + - 1 + - 2 + - 3 + allow_generated_as_anchor: false + diverse_selection: true + compress: + downsample_ratio: 2 + dynamic: + enabled: true + recent_frames: 4 + conv_kernel_t: 3 + conv_stride_t: 2 + compress: + downsample_ratio: 2 + revisit: + enabled: true + deterministic_pose_retrieval: true + fov_overlap_threshold: 0.6 + plucker_weight: 0.1 + max_frames: 2 + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + compress: + downsample_ratio: 2 + eval_ablation: + enabled: false + branch: A_plus_D_plus_R_normal + generated_history_proxy: + enabled: true + start_step: 40000 + ramp_steps: 40000 + max_prob: 0.25 + noise_std: 0.25 + dropout_prob: 0.0 + injection: + memory_hidden_size: 512 + cache: + enabled: true + device: cpu + keep_raw_latents: all + keep_compressed_records: true + keep_prefix_anchors: true + eviction_policy: none + no_evict: true + clear_between_videos: true + max_records: null + on_capacity_exceeded: warn + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 4 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/18-56-26/.hydra/hydra.yaml b/outputs/2026-06-19/18-56-26/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cb40de80e346caa3e9f7df77b383a6642f9e8f3c --- /dev/null +++ b/outputs/2026-06-19/18-56-26/.hydra/hydra.yaml @@ -0,0 +1,232 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_memory_dit + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - +dataset.customized_validation=true + - +dataset.wo_updown=false + - +dataset.angle_range=180 + - +dataset.pos_range=8 + - ++algorithm.n_tokens=4 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - algorithm.memory_cross_attn_layers=[3,7,11,15] + - ++algorithm.dememwm.injection.memory_hidden_size=512 + - ++algorithm.dememwm.generated_history_proxy.enabled=true + - ++algorithm.dememwm.generated_history_proxy.start_step=40000 + - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 + - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 + - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 + - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 + - ++algorithm.dememwm.anchor.enabled=true + - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] + - ++algorithm.dememwm.anchor.diverse_selection=true + - ++algorithm.dememwm.anchor.compress.downsample_ratio=2 + - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false + - ++algorithm.dememwm.dynamic.enabled=true + - ++algorithm.dememwm.dynamic.recent_frames=4 + - ++algorithm.dememwm.dynamic.conv_stride_t=2 + - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 + - ++algorithm.dememwm.revisit.enabled=true + - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true + - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.max_frames=2 + - ++algorithm.dememwm.revisit.compress.downsample_ratio=2 + - ++algorithm.dememwm.cache.enabled=true + - ++algorithm.dememwm.cache.device=cpu + - ++algorithm.dememwm.cache.keep_raw_latents=all + - ++algorithm.dememwm.cache.keep_compressed_records=true + - ++algorithm.dememwm.cache.eviction_policy=none + - ++algorithm.dememwm.cache.no_evict=true + - ++algorithm.dememwm.cache.clear_between_videos=true + - ++algorithm.dememwm.cache.max_records=null + - ++algorithm.dememwm.cache.on_capacity_exceeded=warn + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=20000 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.anchor.allow_generated_as_anchor=false,++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3],++algorithm.dememwm.anchor.compress.downsample_ratio=2,++algorithm.dememwm.anchor.diverse_selection=true,++algorithm.dememwm.anchor.enabled=true,++algorithm.dememwm.cache.clear_between_videos=true,++algorithm.dememwm.cache.device=cpu,++algorithm.dememwm.cache.enabled=true,++algorithm.dememwm.cache.eviction_policy=none,++algorithm.dememwm.cache.keep_compressed_records=true,++algorithm.dememwm.cache.keep_raw_latents=all,++algorithm.dememwm.cache.max_records=null,++algorithm.dememwm.cache.no_evict=true,++algorithm.dememwm.cache.on_capacity_exceeded=warn,++algorithm.dememwm.curriculum.dit_freeze.enabled=true,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=20000,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.dynamic.compress.downsample_ratio=2,++algorithm.dememwm.dynamic.conv_stride_t=2,++algorithm.dememwm.dynamic.enabled=true,++algorithm.dememwm.dynamic.recent_frames=4,++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0,++algorithm.dememwm.generated_history_proxy.enabled=true,++algorithm.dememwm.generated_history_proxy.max_prob=0.25,++algorithm.dememwm.generated_history_proxy.noise_std=0.25,++algorithm.dememwm.generated_history_proxy.ramp_steps=40000,++algorithm.dememwm.generated_history_proxy.start_step=40000,++algorithm.dememwm.injection.memory_hidden_size=512,++algorithm.dememwm.revisit.compress.downsample_ratio=2,++algorithm.dememwm.revisit.deterministic_pose_retrieval=true,++algorithm.dememwm.revisit.enabled=true,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_overlap_threshold=0.60,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.max_frames=2,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=4,+customized_load=true,+dataset.angle_range=180,+dataset.customized_validation=true,+dataset.n_frames_valid=500,+dataset.pos_range=8,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.memory_cross_attn_layers=[3,7,11,15],algorithm.x_shape=[16,18,32],algorithm=dememwm_memory_dit,auto_resume=true,dataset.n_frames=1000,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/18-56-26 + choices: + cluster: null + algorithm: dememwm_memory_dit + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/18-56-26/.hydra/overrides.yaml b/outputs/2026-06-19/18-56-26/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8ee3288e706e04827ca476ead0592fea9c288ffa --- /dev/null +++ b/outputs/2026-06-19/18-56-26/.hydra/overrides.yaml @@ -0,0 +1,74 @@ +- +name=train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_memory_dit +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- +dataset.customized_validation=true +- +dataset.wo_updown=false +- +dataset.angle_range=180 +- +dataset.pos_range=8 +- ++algorithm.n_tokens=4 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- algorithm.memory_cross_attn_layers=[3,7,11,15] +- ++algorithm.dememwm.injection.memory_hidden_size=512 +- ++algorithm.dememwm.generated_history_proxy.enabled=true +- ++algorithm.dememwm.generated_history_proxy.start_step=40000 +- ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000 +- ++algorithm.dememwm.generated_history_proxy.max_prob=0.25 +- ++algorithm.dememwm.generated_history_proxy.noise_std=0.25 +- ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0 +- ++algorithm.dememwm.anchor.enabled=true +- ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3] +- ++algorithm.dememwm.anchor.diverse_selection=true +- ++algorithm.dememwm.anchor.compress.downsample_ratio=2 +- ++algorithm.dememwm.anchor.allow_generated_as_anchor=false +- ++algorithm.dememwm.dynamic.enabled=true +- ++algorithm.dememwm.dynamic.recent_frames=4 +- ++algorithm.dememwm.dynamic.conv_stride_t=2 +- ++algorithm.dememwm.dynamic.compress.downsample_ratio=2 +- ++algorithm.dememwm.revisit.enabled=true +- ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true +- ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.max_frames=2 +- ++algorithm.dememwm.revisit.compress.downsample_ratio=2 +- ++algorithm.dememwm.cache.enabled=true +- ++algorithm.dememwm.cache.device=cpu +- ++algorithm.dememwm.cache.keep_raw_latents=all +- ++algorithm.dememwm.cache.keep_compressed_records=true +- ++algorithm.dememwm.cache.eviction_policy=none +- ++algorithm.dememwm.cache.no_evict=true +- ++algorithm.dememwm.cache.clear_between_videos=true +- ++algorithm.dememwm.cache.max_records=null +- ++algorithm.dememwm.cache.on_capacity_exceeded=warn +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=20000 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.dit_freeze.enabled=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-19/18-56-26/main.log b/outputs/2026-06-19/18-56-26/main.log new file mode 100644 index 0000000000000000000000000000000000000000..4df50e7e80675c88f553b0c00b1ff48df43f9df2 --- /dev/null +++ b/outputs/2026-06-19/18-56-26/main.log @@ -0,0 +1,48 @@ +[2026-06-19 18:57:04,868][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-19 18:57:05,006][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-19 18:57:05,006][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-19 18:57:05,006][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-19 18:57:05,006][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-19 18:57:05,007][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-19 18:57:06,925][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-19 18:57:06,926][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-19 18:57:06,942][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-19 18:57:06,949][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-19 18:57:12,511][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-19 18:57:30,579][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/checkpoints/last.ckpt +[2026-06-19 18:57:47,228][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 18:57:47,230][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 18:57:47,230][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 18:57:47,344][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 18:57:48,995][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +------------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 681 M +1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 887 K +2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 296 K +3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 296 K +4 | dememwm_revisit_gate | RevisitRawGate | 4 +5 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +6 | vae | AutoencoderKL | 229 M +------------------------------------------------------------------------------------- +682 M Trainable params +231 M Non-trainable params +914 M Total params +3,658.543 Total estimated model params size (MB) +[2026-06-19 18:57:49,674][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_separate_memory_gatefix_4layers512_valid500_berzelius_8a100_bs16_global128_350k/checkpoints/last.ckpt +[2026-06-19 18:57:52,222][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 18:57:52,222][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 18:57:52,222][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 18:57:52,223][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 19:04:55,614][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-19 19:04:55,614][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-19 19:04:55,614][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-19 19:04:55,613][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-19 19:04:55,613][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-19 19:04:55,614][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-19 19:04:55,614][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-19 19:04:55,614][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-19/19-05-08/.hydra/config.yaml b/outputs/2026-06-19/19-05-08/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c59a632cf78c535166e69ca7f9d7cc18ae1e3f3b --- /dev/null +++ b/outputs/2026-06-19/19-05-08/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-05-08/.hydra/hydra.yaml b/outputs/2026-06-19/19-05-08/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..35ccf34cb07324f71c92458d94078f6924679b77 --- /dev/null +++ b/outputs/2026-06-19/19-05-08/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-05-08 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-05-08/.hydra/overrides.yaml b/outputs/2026-06-19/19-05-08/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7dc9321e7110bcf1e972bf4ff4e64f393e83fa21 --- /dev/null +++ b/outputs/2026-06-19/19-05-08/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-05-08/main.log b/outputs/2026-06-19/19-05-08/main.log new file mode 100644 index 0000000000000000000000000000000000000000..11c24081dba1be91ef7588f0dc994871d422f054 --- /dev/null +++ b/outputs/2026-06-19/19-05-08/main.log @@ -0,0 +1 @@ +[2026-06-19 19:06:02,006][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 diff --git a/outputs/2026-06-19/19-05-09/.hydra/config.yaml b/outputs/2026-06-19/19-05-09/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c59a632cf78c535166e69ca7f9d7cc18ae1e3f3b --- /dev/null +++ b/outputs/2026-06-19/19-05-09/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-05-09/.hydra/hydra.yaml b/outputs/2026-06-19/19-05-09/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..21b6a244a54a15710c68ecb8199e8dc7aebf189f --- /dev/null +++ b/outputs/2026-06-19/19-05-09/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-05-09 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-05-09/.hydra/overrides.yaml b/outputs/2026-06-19/19-05-09/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7dc9321e7110bcf1e972bf4ff4e64f393e83fa21 --- /dev/null +++ b/outputs/2026-06-19/19-05-09/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-05-09/main.log b/outputs/2026-06-19/19-05-09/main.log new file mode 100644 index 0000000000000000000000000000000000000000..a90452ca5bc0e971c6383abd05fb68f13a820e77 --- /dev/null +++ b/outputs/2026-06-19/19-05-09/main.log @@ -0,0 +1,14 @@ +[2026-06-19 19:05:55,848][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-19 19:05:56,008][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-19 19:05:56,009][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-19 19:05:56,009][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-19 19:05:56,009][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-19 19:05:56,009][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-19 19:05:58,400][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-19 19:05:59,667][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-19 19:06:02,045][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-19 19:06:09,287][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + diff --git a/outputs/2026-06-19/19-05-12/.hydra/config.yaml b/outputs/2026-06-19/19-05-12/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c59a632cf78c535166e69ca7f9d7cc18ae1e3f3b --- /dev/null +++ b/outputs/2026-06-19/19-05-12/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-05-12/.hydra/hydra.yaml b/outputs/2026-06-19/19-05-12/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..2b6d5146597d7abef75cfc44d2c51863939cfb36 --- /dev/null +++ b/outputs/2026-06-19/19-05-12/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-05-12 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-05-12/.hydra/overrides.yaml b/outputs/2026-06-19/19-05-12/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7dc9321e7110bcf1e972bf4ff4e64f393e83fa21 --- /dev/null +++ b/outputs/2026-06-19/19-05-12/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-05-12/main.log b/outputs/2026-06-19/19-05-12/main.log new file mode 100644 index 0000000000000000000000000000000000000000..d7ed6ed3c73a349a204b0a811e915d68481732a0 --- /dev/null +++ b/outputs/2026-06-19/19-05-12/main.log @@ -0,0 +1,2 @@ +[2026-06-19 19:06:02,001][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-19 19:06:02,006][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 diff --git a/outputs/2026-06-19/19-05-13/.hydra/config.yaml b/outputs/2026-06-19/19-05-13/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c59a632cf78c535166e69ca7f9d7cc18ae1e3f3b --- /dev/null +++ b/outputs/2026-06-19/19-05-13/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-05-13/.hydra/hydra.yaml b/outputs/2026-06-19/19-05-13/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..54e36110b69b5a645f9ec88c5da7c844b8e4f4d7 --- /dev/null +++ b/outputs/2026-06-19/19-05-13/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-05-13 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-05-13/.hydra/overrides.yaml b/outputs/2026-06-19/19-05-13/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7dc9321e7110bcf1e972bf4ff4e64f393e83fa21 --- /dev/null +++ b/outputs/2026-06-19/19-05-13/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-05-13/main.log b/outputs/2026-06-19/19-05-13/main.log new file mode 100644 index 0000000000000000000000000000000000000000..566749444d3baf62cf5b95f67c7d15f085d71346 --- /dev/null +++ b/outputs/2026-06-19/19-05-13/main.log @@ -0,0 +1,2 @@ +[2026-06-19 19:06:02,002][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-19 19:06:02,031][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 diff --git a/outputs/2026-06-19/19-05-14/.hydra/config.yaml b/outputs/2026-06-19/19-05-14/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c59a632cf78c535166e69ca7f9d7cc18ae1e3f3b --- /dev/null +++ b/outputs/2026-06-19/19-05-14/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-05-14/.hydra/hydra.yaml b/outputs/2026-06-19/19-05-14/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4f1b62e6b712abdf2eb4fd160d9ed74e82e1eb72 --- /dev/null +++ b/outputs/2026-06-19/19-05-14/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-05-14 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-05-14/.hydra/overrides.yaml b/outputs/2026-06-19/19-05-14/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7dc9321e7110bcf1e972bf4ff4e64f393e83fa21 --- /dev/null +++ b/outputs/2026-06-19/19-05-14/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-05-14/main.log b/outputs/2026-06-19/19-05-14/main.log new file mode 100644 index 0000000000000000000000000000000000000000..efb1add8f763818cb29235a2d8425d1e9ca5f68b --- /dev/null +++ b/outputs/2026-06-19/19-05-14/main.log @@ -0,0 +1 @@ +[2026-06-19 19:06:02,012][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 diff --git a/outputs/2026-06-19/19-05-15/.hydra/config.yaml b/outputs/2026-06-19/19-05-15/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c59a632cf78c535166e69ca7f9d7cc18ae1e3f3b --- /dev/null +++ b/outputs/2026-06-19/19-05-15/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-05-15/.hydra/hydra.yaml b/outputs/2026-06-19/19-05-15/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c8fa99af23acfa34c271ae7a716546d2c85567df --- /dev/null +++ b/outputs/2026-06-19/19-05-15/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-05-15 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-05-15/.hydra/overrides.yaml b/outputs/2026-06-19/19-05-15/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7dc9321e7110bcf1e972bf4ff4e64f393e83fa21 --- /dev/null +++ b/outputs/2026-06-19/19-05-15/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-05-15/main.log b/outputs/2026-06-19/19-05-15/main.log new file mode 100644 index 0000000000000000000000000000000000000000..a0e87af8f98e6e6afaf58b74cfe0f0b18a6e8909 --- /dev/null +++ b/outputs/2026-06-19/19-05-15/main.log @@ -0,0 +1 @@ +[2026-06-19 19:06:02,001][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 diff --git a/outputs/2026-06-19/19-06-42/.hydra/config.yaml b/outputs/2026-06-19/19-06-42/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/19-06-42/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-06-42/.hydra/hydra.yaml b/outputs/2026-06-19/19-06-42/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..855c5c92bc6723d8340ff86ea468c4b9dd8d68d6 --- /dev/null +++ b/outputs/2026-06-19/19-06-42/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-06-42 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-06-42/.hydra/overrides.yaml b/outputs/2026-06-19/19-06-42/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/19-06-42/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-06-42/main.log b/outputs/2026-06-19/19-06-42/main.log new file mode 100644 index 0000000000000000000000000000000000000000..65d98268ccd1718bfb4637845a1839165fa01541 --- /dev/null +++ b/outputs/2026-06-19/19-06-42/main.log @@ -0,0 +1 @@ +[2026-06-19 19:07:19,033][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 diff --git a/outputs/2026-06-19/19-06-43/.hydra/config.yaml b/outputs/2026-06-19/19-06-43/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/19-06-43/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-06-43/.hydra/hydra.yaml b/outputs/2026-06-19/19-06-43/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..2834521d34683d9089c24181b4a1ea53041686f2 --- /dev/null +++ b/outputs/2026-06-19/19-06-43/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-06-43 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-06-43/.hydra/overrides.yaml b/outputs/2026-06-19/19-06-43/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/19-06-43/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-06-43/main.log b/outputs/2026-06-19/19-06-43/main.log new file mode 100644 index 0000000000000000000000000000000000000000..c273f1cf389c6154046de7f636e8a2b778ab5db5 --- /dev/null +++ b/outputs/2026-06-19/19-06-43/main.log @@ -0,0 +1,2 @@ +[2026-06-19 19:07:18,962][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-19 19:07:19,021][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 diff --git a/outputs/2026-06-19/19-06-44/.hydra/config.yaml b/outputs/2026-06-19/19-06-44/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/19-06-44/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-06-44/.hydra/hydra.yaml b/outputs/2026-06-19/19-06-44/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9fbd52f1796a656f8f22b5150c44e17b753d5d1b --- /dev/null +++ b/outputs/2026-06-19/19-06-44/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-06-44 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-06-44/.hydra/overrides.yaml b/outputs/2026-06-19/19-06-44/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/19-06-44/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-06-44/main.log b/outputs/2026-06-19/19-06-44/main.log new file mode 100644 index 0000000000000000000000000000000000000000..ecda8bbb07aaeedc8f552d47a8d4c6ee4106908f --- /dev/null +++ b/outputs/2026-06-19/19-06-44/main.log @@ -0,0 +1,14 @@ +[2026-06-19 19:07:14,036][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-19 19:07:14,176][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-19 19:07:14,177][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-19 19:07:14,177][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-19 19:07:14,177][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-19 19:07:14,177][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-19 19:07:16,054][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-19 19:07:17,372][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-19 19:07:18,994][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-19 19:07:25,541][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + diff --git a/outputs/2026-06-19/19-06-45/.hydra/config.yaml b/outputs/2026-06-19/19-06-45/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/19-06-45/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-06-45/.hydra/hydra.yaml b/outputs/2026-06-19/19-06-45/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ab52de1e296a483ae3adfdc70c72f9075638c8b2 --- /dev/null +++ b/outputs/2026-06-19/19-06-45/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-06-45 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-06-45/.hydra/overrides.yaml b/outputs/2026-06-19/19-06-45/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/19-06-45/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-06-45/main.log b/outputs/2026-06-19/19-06-45/main.log new file mode 100644 index 0000000000000000000000000000000000000000..63b54aa679fb57938001369b4f0b0c5fba13c09c --- /dev/null +++ b/outputs/2026-06-19/19-06-45/main.log @@ -0,0 +1 @@ +[2026-06-19 19:07:18,963][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 diff --git a/outputs/2026-06-19/19-06-46/.hydra/config.yaml b/outputs/2026-06-19/19-06-46/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/19-06-46/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-06-46/.hydra/hydra.yaml b/outputs/2026-06-19/19-06-46/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..57e81a930efd6a868e26732080eb69b682da7a58 --- /dev/null +++ b/outputs/2026-06-19/19-06-46/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-06-46 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-06-46/.hydra/overrides.yaml b/outputs/2026-06-19/19-06-46/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/19-06-46/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-06-46/main.log b/outputs/2026-06-19/19-06-46/main.log new file mode 100644 index 0000000000000000000000000000000000000000..7bb7de78b1e7da924c1e6068e459dbd8d113797d --- /dev/null +++ b/outputs/2026-06-19/19-06-46/main.log @@ -0,0 +1 @@ +[2026-06-19 19:07:18,972][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 diff --git a/outputs/2026-06-19/19-06-47/.hydra/config.yaml b/outputs/2026-06-19/19-06-47/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/19-06-47/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-06-47/.hydra/hydra.yaml b/outputs/2026-06-19/19-06-47/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..945b02f52bd8f707df8247c3e001e89a4fd105e2 --- /dev/null +++ b/outputs/2026-06-19/19-06-47/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-06-47 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-06-47/.hydra/overrides.yaml b/outputs/2026-06-19/19-06-47/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/19-06-47/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-06-47/main.log b/outputs/2026-06-19/19-06-47/main.log new file mode 100644 index 0000000000000000000000000000000000000000..6716674bc4ee5ed76b5b02d12ebbb150adb79974 --- /dev/null +++ b/outputs/2026-06-19/19-06-47/main.log @@ -0,0 +1 @@ +[2026-06-19 19:07:18,958][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 diff --git a/outputs/2026-06-19/19-06-48/.hydra/config.yaml b/outputs/2026-06-19/19-06-48/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/19-06-48/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-06-48/.hydra/hydra.yaml b/outputs/2026-06-19/19-06-48/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b10ca39e12de57c870f44971a0d42c2d07d0bd0d --- /dev/null +++ b/outputs/2026-06-19/19-06-48/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-06-48 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-06-48/.hydra/overrides.yaml b/outputs/2026-06-19/19-06-48/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/19-06-48/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-06-48/main.log b/outputs/2026-06-19/19-06-48/main.log new file mode 100644 index 0000000000000000000000000000000000000000..efedf29d24756eb917018c4f92017e1e92af1c1c --- /dev/null +++ b/outputs/2026-06-19/19-06-48/main.log @@ -0,0 +1 @@ +[2026-06-19 19:07:18,993][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 diff --git a/outputs/2026-06-19/19-07-41/.hydra/config.yaml b/outputs/2026-06-19/19-07-41/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/19-07-41/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-07-41/.hydra/hydra.yaml b/outputs/2026-06-19/19-07-41/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e534c6114b40af3363b58e2895e9b81c62e41af6 --- /dev/null +++ b/outputs/2026-06-19/19-07-41/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-07-41 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-07-41/.hydra/overrides.yaml b/outputs/2026-06-19/19-07-41/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/19-07-41/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-07-41/main.log b/outputs/2026-06-19/19-07-41/main.log new file mode 100644 index 0000000000000000000000000000000000000000..a705e6df481d9f4c13201aa3672d4a2bb22d5560 --- /dev/null +++ b/outputs/2026-06-19/19-07-41/main.log @@ -0,0 +1 @@ +[2026-06-19 19:08:17,014][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 diff --git a/outputs/2026-06-19/19-07-42/.hydra/config.yaml b/outputs/2026-06-19/19-07-42/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/19-07-42/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-07-42/.hydra/hydra.yaml b/outputs/2026-06-19/19-07-42/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7037f8f2eda344b1870b3456a71db4678ae2d21f --- /dev/null +++ b/outputs/2026-06-19/19-07-42/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-07-42 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-07-42/.hydra/overrides.yaml b/outputs/2026-06-19/19-07-42/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/19-07-42/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-07-42/main.log b/outputs/2026-06-19/19-07-42/main.log new file mode 100644 index 0000000000000000000000000000000000000000..25856a619345bdad65760ea9e25c0f0d4059a957 --- /dev/null +++ b/outputs/2026-06-19/19-07-42/main.log @@ -0,0 +1,2 @@ +[2026-06-19 19:08:15,904][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-19 19:08:16,370][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 diff --git a/outputs/2026-06-19/19-07-43/.hydra/config.yaml b/outputs/2026-06-19/19-07-43/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/19-07-43/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-07-43/.hydra/hydra.yaml b/outputs/2026-06-19/19-07-43/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d1049a00b73ab5c5312c440aa5863bbe8ebfbfff --- /dev/null +++ b/outputs/2026-06-19/19-07-43/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-07-43 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-07-43/.hydra/overrides.yaml b/outputs/2026-06-19/19-07-43/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/19-07-43/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-07-43/main.log b/outputs/2026-06-19/19-07-43/main.log new file mode 100644 index 0000000000000000000000000000000000000000..60e014fca511b23b2b9b70b050bcfcd3e261b79d --- /dev/null +++ b/outputs/2026-06-19/19-07-43/main.log @@ -0,0 +1,17 @@ +[2026-06-19 19:08:14,332][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-19 19:08:15,193][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-19 19:08:15,193][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-19 19:08:15,193][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-19 19:08:15,193][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-19 19:08:15,194][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-19 19:08:15,299][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-19 19:08:15,397][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-19 19:08:16,134][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-19 19:08:16,172][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-19 19:08:16,867][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-19 19:08:17,015][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-19 19:08:22,716][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + diff --git a/outputs/2026-06-19/19-07-44/.hydra/config.yaml b/outputs/2026-06-19/19-07-44/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..123262f53d3a02bc01b30dd13312424c4c28d9c6 --- /dev/null +++ b/outputs/2026-06-19/19-07-44/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: online +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-07-44/.hydra/hydra.yaml b/outputs/2026-06-19/19-07-44/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c73577a85e32c4cc1fcd6f16a896300059d02b67 --- /dev/null +++ b/outputs/2026-06-19/19-07-44/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - wandb.mode=online + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=online + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-07-44 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-07-44/.hydra/overrides.yaml b/outputs/2026-06-19/19-07-44/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75166170fa8010ddd694dc7475a3046fac0c6f60 --- /dev/null +++ b/outputs/2026-06-19/19-07-44/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- wandb.mode=online +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-07-44/main.log b/outputs/2026-06-19/19-07-44/main.log new file mode 100644 index 0000000000000000000000000000000000000000..f035b3c9e096a69b9d7cd63dc8316d413e1bec5a --- /dev/null +++ b/outputs/2026-06-19/19-07-44/main.log @@ -0,0 +1 @@ +[2026-06-19 19:08:16,642][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 diff --git a/outputs/2026-06-19/19-13-42/.hydra/config.yaml b/outputs/2026-06-19/19-13-42/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..dfda8864731f92d059d677bea2ca21a466ac793c --- /dev/null +++ b/outputs/2026-06-19/19-13-42/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-13-42/.hydra/hydra.yaml b/outputs/2026-06-19/19-13-42/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ced671ec869c943d624769e6482972cb1f27d06b --- /dev/null +++ b/outputs/2026-06-19/19-13-42/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-13-42 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-13-42/.hydra/overrides.yaml b/outputs/2026-06-19/19-13-42/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0161cbdaa34e92ea4a75e133231c3c78fc6c89af --- /dev/null +++ b/outputs/2026-06-19/19-13-42/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-13-42/main.log b/outputs/2026-06-19/19-13-42/main.log new file mode 100644 index 0000000000000000000000000000000000000000..b7fa6138d5f569fab25ae79f2dcec75ff40dc3de --- /dev/null +++ b/outputs/2026-06-19/19-13-42/main.log @@ -0,0 +1,6 @@ +[2026-06-19 19:14:16,778][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-19 19:14:17,512][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-19 19:14:40,139][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 19:14:40,139][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 19:14:42,940][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 19:14:42,940][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-19/19-13-43/.hydra/config.yaml b/outputs/2026-06-19/19-13-43/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..dfda8864731f92d059d677bea2ca21a466ac793c --- /dev/null +++ b/outputs/2026-06-19/19-13-43/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-13-43/.hydra/hydra.yaml b/outputs/2026-06-19/19-13-43/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a5742015f3712ac19e39d4e2250da75114636652 --- /dev/null +++ b/outputs/2026-06-19/19-13-43/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-13-43 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-13-43/.hydra/overrides.yaml b/outputs/2026-06-19/19-13-43/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0161cbdaa34e92ea4a75e133231c3c78fc6c89af --- /dev/null +++ b/outputs/2026-06-19/19-13-43/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-13-43/main.log b/outputs/2026-06-19/19-13-43/main.log new file mode 100644 index 0000000000000000000000000000000000000000..9af1185852f46a93e2441b83ae7b8c2260a1f52c --- /dev/null +++ b/outputs/2026-06-19/19-13-43/main.log @@ -0,0 +1,6 @@ +[2026-06-19 19:14:16,858][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-19 19:14:16,906][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-19 19:14:40,139][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 19:14:40,139][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 19:14:42,940][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 19:14:42,940][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-19/19-13-44/.hydra/config.yaml b/outputs/2026-06-19/19-13-44/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..dfda8864731f92d059d677bea2ca21a466ac793c --- /dev/null +++ b/outputs/2026-06-19/19-13-44/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-13-44/.hydra/hydra.yaml b/outputs/2026-06-19/19-13-44/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cc699184d7b57de5292515eb9b4e18d339513dda --- /dev/null +++ b/outputs/2026-06-19/19-13-44/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-13-44 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-13-44/.hydra/overrides.yaml b/outputs/2026-06-19/19-13-44/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0161cbdaa34e92ea4a75e133231c3c78fc6c89af --- /dev/null +++ b/outputs/2026-06-19/19-13-44/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-13-44/main.log b/outputs/2026-06-19/19-13-44/main.log new file mode 100644 index 0000000000000000000000000000000000000000..241a0cf4709e30c29fa588095b9a58c925660a52 --- /dev/null +++ b/outputs/2026-06-19/19-13-44/main.log @@ -0,0 +1,6 @@ +[2026-06-19 19:14:16,913][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-19 19:14:17,667][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-19 19:14:40,139][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 19:14:40,139][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 19:14:42,940][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 19:14:42,940][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-19/19-13-45/.hydra/config.yaml b/outputs/2026-06-19/19-13-45/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..dfda8864731f92d059d677bea2ca21a466ac793c --- /dev/null +++ b/outputs/2026-06-19/19-13-45/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-13-45/.hydra/hydra.yaml b/outputs/2026-06-19/19-13-45/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..27ede79d9224b627ab11055c34164b1d7cf2dcb8 --- /dev/null +++ b/outputs/2026-06-19/19-13-45/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-13-45 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-13-45/.hydra/overrides.yaml b/outputs/2026-06-19/19-13-45/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0161cbdaa34e92ea4a75e133231c3c78fc6c89af --- /dev/null +++ b/outputs/2026-06-19/19-13-45/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_a2d4r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-13-45/main.log b/outputs/2026-06-19/19-13-45/main.log new file mode 100644 index 0000000000000000000000000000000000000000..9eed4d431638a5d715f7d04d0efd4a939721e1f5 --- /dev/null +++ b/outputs/2026-06-19/19-13-45/main.log @@ -0,0 +1,3 @@ +[2026-06-19 19:14:16,693][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-19 19:14:40,139][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 19:14:42,939][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-19/19-13-46/.hydra/config.yaml b/outputs/2026-06-19/19-13-46/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5deab422580abeed5d76ece0f601c27f3c6f094d --- /dev/null +++ b/outputs/2026-06-19/19-13-46/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-13-46/.hydra/hydra.yaml b/outputs/2026-06-19/19-13-46/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7400cf4ee0dc1966e12772b4b8377a2dd34f71fd --- /dev/null +++ b/outputs/2026-06-19/19-13-46/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-13-46 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-13-46/.hydra/overrides.yaml b/outputs/2026-06-19/19-13-46/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b3d602b2f3f21e413975c8c5b2dc7365caa1d6d3 --- /dev/null +++ b/outputs/2026-06-19/19-13-46/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-13-46/main.log b/outputs/2026-06-19/19-13-46/main.log new file mode 100644 index 0000000000000000000000000000000000000000..b25ec9d42058db9b303d83751ef1e82103586170 Binary files /dev/null and b/outputs/2026-06-19/19-13-46/main.log differ diff --git a/outputs/2026-06-19/19-13-47/.hydra/config.yaml b/outputs/2026-06-19/19-13-47/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5deab422580abeed5d76ece0f601c27f3c6f094d --- /dev/null +++ b/outputs/2026-06-19/19-13-47/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-13-47/.hydra/hydra.yaml b/outputs/2026-06-19/19-13-47/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..101b48b9f2d33b73a4dfb36d3344e8d94faa0feb --- /dev/null +++ b/outputs/2026-06-19/19-13-47/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-13-47 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-13-47/.hydra/overrides.yaml b/outputs/2026-06-19/19-13-47/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b3d602b2f3f21e413975c8c5b2dc7365caa1d6d3 --- /dev/null +++ b/outputs/2026-06-19/19-13-47/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-13-47/main.log b/outputs/2026-06-19/19-13-47/main.log new file mode 100644 index 0000000000000000000000000000000000000000..2759a9f3f62cc623fff86e40edd3798af03f0076 --- /dev/null +++ b/outputs/2026-06-19/19-13-47/main.log @@ -0,0 +1,31 @@ +[2026-06-19 19:14:19,872][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-19 19:14:20,017][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-19 19:14:20,018][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-19 19:14:20,018][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-19 19:14:20,018][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-19 19:14:20,018][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-19 19:14:20,139][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-19 19:14:20,262][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-19 19:14:21,926][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-19 19:14:22,502][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-19 19:14:28,347][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-19 19:14:46,331][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 19:14:46,331][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 19:14:47,865][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +1.0 B Trainable params +231 M Non-trainable params +1.3 B Total params +5,107.144 Total estimated model params size (MB) +[2026-06-19 19:14:49,766][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-19 19:14:49,766][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 04:51:09,030][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=175000` reached. diff --git a/outputs/2026-06-19/19-13-48/.hydra/config.yaml b/outputs/2026-06-19/19-13-48/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5deab422580abeed5d76ece0f601c27f3c6f094d --- /dev/null +++ b/outputs/2026-06-19/19-13-48/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-13-48/.hydra/hydra.yaml b/outputs/2026-06-19/19-13-48/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e2b374663d6536b0eaa8ea3e4647c67387a2a917 --- /dev/null +++ b/outputs/2026-06-19/19-13-48/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-13-48 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-13-48/.hydra/overrides.yaml b/outputs/2026-06-19/19-13-48/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b3d602b2f3f21e413975c8c5b2dc7365caa1d6d3 --- /dev/null +++ b/outputs/2026-06-19/19-13-48/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-13-48/main.log b/outputs/2026-06-19/19-13-48/main.log new file mode 100644 index 0000000000000000000000000000000000000000..7e29bb92ca90dafe12df753e16355827c1825eb8 --- /dev/null +++ b/outputs/2026-06-19/19-13-48/main.log @@ -0,0 +1,3 @@ +[2026-06-19 19:14:21,858][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-19 19:14:46,331][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 19:14:49,766][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-19/19-13-49/.hydra/config.yaml b/outputs/2026-06-19/19-13-49/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5deab422580abeed5d76ece0f601c27f3c6f094d --- /dev/null +++ b/outputs/2026-06-19/19-13-49/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-13-49/.hydra/hydra.yaml b/outputs/2026-06-19/19-13-49/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..1bf7253ffd6a5ff25c337048256e1462c9e0f861 --- /dev/null +++ b/outputs/2026-06-19/19-13-49/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-13-49 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-13-49/.hydra/overrides.yaml b/outputs/2026-06-19/19-13-49/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b3d602b2f3f21e413975c8c5b2dc7365caa1d6d3 --- /dev/null +++ b/outputs/2026-06-19/19-13-49/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-13-49/main.log b/outputs/2026-06-19/19-13-49/main.log new file mode 100644 index 0000000000000000000000000000000000000000..380ccad5f11aa594bdd828e0ce5f4d6878df5f73 --- /dev/null +++ b/outputs/2026-06-19/19-13-49/main.log @@ -0,0 +1,3 @@ +[2026-06-19 19:14:21,872][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-19 19:14:46,331][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 19:14:49,766][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-19/19-13-50/.hydra/config.yaml b/outputs/2026-06-19/19-13-50/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5deab422580abeed5d76ece0f601c27f3c6f094d --- /dev/null +++ b/outputs/2026-06-19/19-13-50/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-19/19-13-50/.hydra/hydra.yaml b/outputs/2026-06-19/19-13-50/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..380cdf6ddad3a3324867b3dd44224af67e9a74c4 --- /dev/null +++ b/outputs/2026-06-19/19-13-50/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-19/19-13-50 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-19/19-13-50/.hydra/overrides.yaml b/outputs/2026-06-19/19-13-50/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b3d602b2f3f21e413975c8c5b2dc7365caa1d6d3 --- /dev/null +++ b/outputs/2026-06-19/19-13-50/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_base_a2d4r2_route_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-19/19-13-50/main.log b/outputs/2026-06-19/19-13-50/main.log new file mode 100644 index 0000000000000000000000000000000000000000..9bbca6e29a2235602e1f69dd397ba6074dd9b622 --- /dev/null +++ b/outputs/2026-06-19/19-13-50/main.log @@ -0,0 +1,3 @@ +[2026-06-19 19:14:22,041][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-19 19:14:46,331][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-19 19:14:49,766][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-20/12-19-07/.hydra/config.yaml b/outputs/2026-06-20/12-19-07/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e20b9506faae86321b20e5ef9bdd6bf0c914b072 --- /dev/null +++ b/outputs/2026-06-20/12-19-07/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: false + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-20/12-19-07/.hydra/hydra.yaml b/outputs/2026-06-20/12-19-07/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d5222854d89c0beb0d69c2808081376b36cfb2cd --- /dev/null +++ b/outputs/2026-06-20/12-19-07/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.use_fused_snr=false + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.diffusion.use_fused_snr=false,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-20/12-19-07 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-20/12-19-07/.hydra/overrides.yaml b/outputs/2026-06-20/12-19-07/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..270d3b697f2e4bc4f82664afd6d5ebcd6d161d0a --- /dev/null +++ b/outputs/2026-06-20/12-19-07/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.use_fused_snr=false +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-20/12-19-07/main.log b/outputs/2026-06-20/12-19-07/main.log new file mode 100644 index 0000000000000000000000000000000000000000..5a266141d66cad05fb52ab8a42d5d370c4dbaf46 --- /dev/null +++ b/outputs/2026-06-20/12-19-07/main.log @@ -0,0 +1,10 @@ +[2026-06-20 12:20:01,739][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-20 12:20:01,747][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-20 12:20:24,697][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 12:20:24,697][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 12:20:31,774][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-20 12:20:31,774][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-23 12:19:16,363][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-06-23 12:19:16,365][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-23 12:19:16,363][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-06-23 12:19:16,366][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-20/12-19-08/.hydra/config.yaml b/outputs/2026-06-20/12-19-08/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e20b9506faae86321b20e5ef9bdd6bf0c914b072 --- /dev/null +++ b/outputs/2026-06-20/12-19-08/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: false + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-20/12-19-08/.hydra/hydra.yaml b/outputs/2026-06-20/12-19-08/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3e95a942c12f1276cbd27800e58c67146dadd3d5 --- /dev/null +++ b/outputs/2026-06-20/12-19-08/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.use_fused_snr=false + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.diffusion.use_fused_snr=false,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-20/12-19-08 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-20/12-19-08/.hydra/overrides.yaml b/outputs/2026-06-20/12-19-08/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..270d3b697f2e4bc4f82664afd6d5ebcd6d161d0a --- /dev/null +++ b/outputs/2026-06-20/12-19-08/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.use_fused_snr=false +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-20/12-19-08/main.log b/outputs/2026-06-20/12-19-08/main.log new file mode 100644 index 0000000000000000000000000000000000000000..fd16e739b82e89cfd25f470415fb0d1f2c302332 --- /dev/null +++ b/outputs/2026-06-20/12-19-08/main.log @@ -0,0 +1,34 @@ +[2026-06-20 12:19:56,037][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-20 12:19:56,142][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-20 12:19:56,142][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-20 12:19:56,142][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-20 12:19:56,142][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-20 12:19:56,143][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-20 12:19:57,784][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-20 12:19:58,763][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-20 12:20:01,755][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-20 12:20:01,770][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-20 12:20:08,050][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-20 12:20:24,697][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 12:20:24,697][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 12:20:26,768][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +1.0 B Trainable params +231 M Non-trainable params +1.3 B Total params +5,107.144 Total estimated model params size (MB) +[2026-06-20 12:20:31,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-20 12:20:31,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-23 12:19:16,364][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-06-23 12:19:16,364][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-06-23 12:19:16,366][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-06-23 12:19:16,366][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-20/12-19-09/.hydra/config.yaml b/outputs/2026-06-20/12-19-09/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e20b9506faae86321b20e5ef9bdd6bf0c914b072 --- /dev/null +++ b/outputs/2026-06-20/12-19-09/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: false + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-20/12-19-09/.hydra/hydra.yaml b/outputs/2026-06-20/12-19-09/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..66bab7c3a1cbc9adf2f272a0d9c9e3230613dbb2 --- /dev/null +++ b/outputs/2026-06-20/12-19-09/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.use_fused_snr=false + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.diffusion.use_fused_snr=false,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-20/12-19-09 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-20/12-19-09/.hydra/overrides.yaml b/outputs/2026-06-20/12-19-09/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..270d3b697f2e4bc4f82664afd6d5ebcd6d161d0a --- /dev/null +++ b/outputs/2026-06-20/12-19-09/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.use_fused_snr=false +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-20/12-19-09/main.log b/outputs/2026-06-20/12-19-09/main.log new file mode 100644 index 0000000000000000000000000000000000000000..8137993111a60f11808558c864bd586424d34b94 --- /dev/null +++ b/outputs/2026-06-20/12-19-09/main.log @@ -0,0 +1,5 @@ +[2026-06-20 12:20:01,763][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-20 12:20:24,697][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 12:20:31,774][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-23 12:19:16,364][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-06-23 12:19:16,366][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-20/12-19-10/.hydra/config.yaml b/outputs/2026-06-20/12-19-10/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e20b9506faae86321b20e5ef9bdd6bf0c914b072 --- /dev/null +++ b/outputs/2026-06-20/12-19-10/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: false + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-20/12-19-10/.hydra/hydra.yaml b/outputs/2026-06-20/12-19-10/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..70a9946891fd79adc4068dffc75b57c831cfe085 --- /dev/null +++ b/outputs/2026-06-20/12-19-10/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.use_fused_snr=false + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.diffusion.use_fused_snr=false,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-20/12-19-10 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-20/12-19-10/.hydra/overrides.yaml b/outputs/2026-06-20/12-19-10/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..270d3b697f2e4bc4f82664afd6d5ebcd6d161d0a --- /dev/null +++ b/outputs/2026-06-20/12-19-10/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.use_fused_snr=false +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-20/12-19-10/main.log b/outputs/2026-06-20/12-19-10/main.log new file mode 100644 index 0000000000000000000000000000000000000000..c76b764d24c5400ac81170aad783eb1c250e3761 --- /dev/null +++ b/outputs/2026-06-20/12-19-10/main.log @@ -0,0 +1,5 @@ +[2026-06-20 12:20:01,758][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-20 12:20:24,697][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 12:20:31,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-23 12:19:16,363][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-06-23 12:19:16,365][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-20/12-19-11/.hydra/config.yaml b/outputs/2026-06-20/12-19-11/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e20b9506faae86321b20e5ef9bdd6bf0c914b072 --- /dev/null +++ b/outputs/2026-06-20/12-19-11/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: false + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-20/12-19-11/.hydra/hydra.yaml b/outputs/2026-06-20/12-19-11/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8c3fdf9da82d68c17a5108bc446ddebd912a6bb3 --- /dev/null +++ b/outputs/2026-06-20/12-19-11/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.use_fused_snr=false + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.diffusion.use_fused_snr=false,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-20/12-19-11 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-20/12-19-11/.hydra/overrides.yaml b/outputs/2026-06-20/12-19-11/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..270d3b697f2e4bc4f82664afd6d5ebcd6d161d0a --- /dev/null +++ b/outputs/2026-06-20/12-19-11/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.use_fused_snr=false +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-20/12-19-11/main.log b/outputs/2026-06-20/12-19-11/main.log new file mode 100644 index 0000000000000000000000000000000000000000..b446545b4010640893e3a257619bbddf6ce64e1f --- /dev/null +++ b/outputs/2026-06-20/12-19-11/main.log @@ -0,0 +1,5 @@ +[2026-06-20 12:20:01,753][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-20 12:20:24,697][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 12:20:31,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-23 12:19:16,365][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-06-23 12:19:16,366][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-20/12-19-12/.hydra/config.yaml b/outputs/2026-06-20/12-19-12/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e20b9506faae86321b20e5ef9bdd6bf0c914b072 --- /dev/null +++ b/outputs/2026-06-20/12-19-12/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 8 + max_epochs: -1 + max_steps: 350000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: false + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 4.0e-05 + memory_adapters: 4.0e-05 + full_dit: 1.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-20/12-19-12/.hydra/hydra.yaml b/outputs/2026-06-20/12-19-12/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a1ac19d5257f184e212e87371a79c7b5c9e170c1 --- /dev/null +++ b/outputs/2026-06-20/12-19-12/.hydra/hydra.yaml @@ -0,0 +1,218 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.use_fused_snr=false + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 + - experiment.training.batch_size=8 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=350000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.diffusion.use_fused_snr=false,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=8,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=350000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-20/12-19-12 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-20/12-19-12/.hydra/overrides.yaml b/outputs/2026-06-20/12-19-12/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..270d3b697f2e4bc4f82664afd6d5ebcd6d161d0a --- /dev/null +++ b/outputs/2026-06-20/12-19-12/.hydra/overrides.yaml @@ -0,0 +1,60 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_snr_shift_a2d4r2_route_ntok8_8a100_bs8_global64_350k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.use_fused_snr=false +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5 +- experiment.training.batch_size=8 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=350000 diff --git a/outputs/2026-06-20/12-19-12/main.log b/outputs/2026-06-20/12-19-12/main.log new file mode 100644 index 0000000000000000000000000000000000000000..1f9781425798c787e961aeb5ca5f22b887ec4f71 --- /dev/null +++ b/outputs/2026-06-20/12-19-12/main.log @@ -0,0 +1,5 @@ +[2026-06-20 12:20:01,739][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-20 12:20:24,697][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 12:20:31,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-23 12:19:16,365][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-06-23 12:19:16,366][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-20/12-58-49/.hydra/config.yaml b/outputs/2026-06-20/12-58-49/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..faf4fdc2cdedc288bdd5309e6df8362385b2f4ea --- /dev/null +++ b/outputs/2026-06-20/12-58-49/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-20/12-58-49/.hydra/hydra.yaml b/outputs/2026-06-20/12-58-49/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0e9c4e5d469397dcff761e49c379737a65467042 --- /dev/null +++ b/outputs/2026-06-20/12-58-49/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-20/12-58-49 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-20/12-58-49/.hydra/overrides.yaml b/outputs/2026-06-20/12-58-49/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9ea76abe63e906884ccf35decf29b5499e51412d --- /dev/null +++ b/outputs/2026-06-20/12-58-49/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-20/12-58-49/main.log b/outputs/2026-06-20/12-58-49/main.log new file mode 100644 index 0000000000000000000000000000000000000000..b1c1faceab1b39385f5ca9436866551d3867d02a --- /dev/null +++ b/outputs/2026-06-20/12-58-49/main.log @@ -0,0 +1,3 @@ +[2026-06-20 12:59:40,037][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-20 13:00:04,596][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 13:00:07,035][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-20/12-58-50/.hydra/config.yaml b/outputs/2026-06-20/12-58-50/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..faf4fdc2cdedc288bdd5309e6df8362385b2f4ea --- /dev/null +++ b/outputs/2026-06-20/12-58-50/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-20/12-58-50/.hydra/hydra.yaml b/outputs/2026-06-20/12-58-50/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a2ae8bdeedbae8d1654cfad41ab4a28dea502844 --- /dev/null +++ b/outputs/2026-06-20/12-58-50/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-20/12-58-50 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-20/12-58-50/.hydra/overrides.yaml b/outputs/2026-06-20/12-58-50/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9ea76abe63e906884ccf35decf29b5499e51412d --- /dev/null +++ b/outputs/2026-06-20/12-58-50/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-20/12-58-50/main.log b/outputs/2026-06-20/12-58-50/main.log new file mode 100644 index 0000000000000000000000000000000000000000..3989993889545f9f177f2be8e8d8041e04a667db --- /dev/null +++ b/outputs/2026-06-20/12-58-50/main.log @@ -0,0 +1,6 @@ +[2026-06-20 12:59:40,037][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-20 12:59:40,065][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-20 13:00:04,596][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 13:00:04,596][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 13:00:07,035][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-20 13:00:07,035][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-20/12-58-51/.hydra/config.yaml b/outputs/2026-06-20/12-58-51/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..faf4fdc2cdedc288bdd5309e6df8362385b2f4ea --- /dev/null +++ b/outputs/2026-06-20/12-58-51/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-20/12-58-51/.hydra/hydra.yaml b/outputs/2026-06-20/12-58-51/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d24648af57a038a1fa9b9df9064704dbe764cbd0 --- /dev/null +++ b/outputs/2026-06-20/12-58-51/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-20/12-58-51 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-20/12-58-51/.hydra/overrides.yaml b/outputs/2026-06-20/12-58-51/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9ea76abe63e906884ccf35decf29b5499e51412d --- /dev/null +++ b/outputs/2026-06-20/12-58-51/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-20/12-58-51/main.log b/outputs/2026-06-20/12-58-51/main.log new file mode 100644 index 0000000000000000000000000000000000000000..d6127192219347fbb38e29ede81d09ea1eaa8291 --- /dev/null +++ b/outputs/2026-06-20/12-58-51/main.log @@ -0,0 +1,3 @@ +[2026-06-20 12:59:40,053][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-20 13:00:04,596][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 13:00:07,035][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-20/12-58-52/.hydra/config.yaml b/outputs/2026-06-20/12-58-52/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..faf4fdc2cdedc288bdd5309e6df8362385b2f4ea --- /dev/null +++ b/outputs/2026-06-20/12-58-52/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-20/12-58-52/.hydra/hydra.yaml b/outputs/2026-06-20/12-58-52/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3acc6009416619afaaaa148576d1b946075d818b --- /dev/null +++ b/outputs/2026-06-20/12-58-52/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-20/12-58-52 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-20/12-58-52/.hydra/overrides.yaml b/outputs/2026-06-20/12-58-52/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9ea76abe63e906884ccf35decf29b5499e51412d --- /dev/null +++ b/outputs/2026-06-20/12-58-52/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-20/12-58-52/main.log b/outputs/2026-06-20/12-58-52/main.log new file mode 100644 index 0000000000000000000000000000000000000000..8e85f54dfb3ad9bf0ca00c735c807c08a4ebe359 --- /dev/null +++ b/outputs/2026-06-20/12-58-52/main.log @@ -0,0 +1,28 @@ +[2026-06-20 12:59:33,702][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-20 12:59:34,857][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-20 12:59:34,857][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-20 12:59:34,857][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-20 12:59:34,857][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-20 12:59:34,858][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-20 12:59:36,657][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-20 12:59:37,984][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-20 12:59:40,037][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-20 12:59:45,715][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-20 13:00:04,596][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 13:00:05,892][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +1.0 B Trainable params +231 M Non-trainable params +1.3 B Total params +5,107.144 Total estimated model params size (MB) +[2026-06-20 13:00:07,035][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 11:10:13,140][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=175000` reached. diff --git a/outputs/2026-06-20/12-58-53/.hydra/config.yaml b/outputs/2026-06-20/12-58-53/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..faf4fdc2cdedc288bdd5309e6df8362385b2f4ea --- /dev/null +++ b/outputs/2026-06-20/12-58-53/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-20/12-58-53/.hydra/hydra.yaml b/outputs/2026-06-20/12-58-53/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e1430318fe06992f5a6e80b17d96553e8f1fabd6 --- /dev/null +++ b/outputs/2026-06-20/12-58-53/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-20/12-58-53 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-20/12-58-53/.hydra/overrides.yaml b/outputs/2026-06-20/12-58-53/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9ea76abe63e906884ccf35decf29b5499e51412d --- /dev/null +++ b/outputs/2026-06-20/12-58-53/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-20/12-58-53/main.log b/outputs/2026-06-20/12-58-53/main.log new file mode 100644 index 0000000000000000000000000000000000000000..a2fe57bed5bf361d592796785a49776f21e9595b --- /dev/null +++ b/outputs/2026-06-20/12-58-53/main.log @@ -0,0 +1,3 @@ +[2026-06-20 12:59:40,051][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-20 13:00:04,596][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 13:00:07,035][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-20/12-58-54/.hydra/config.yaml b/outputs/2026-06-20/12-58-54/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..faf4fdc2cdedc288bdd5309e6df8362385b2f4ea --- /dev/null +++ b/outputs/2026-06-20/12-58-54/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-20/12-58-54/.hydra/hydra.yaml b/outputs/2026-06-20/12-58-54/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0b6e6a42835784bbb44e93f527394a1ad6b4b98f --- /dev/null +++ b/outputs/2026-06-20/12-58-54/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-20/12-58-54 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-20/12-58-54/.hydra/overrides.yaml b/outputs/2026-06-20/12-58-54/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9ea76abe63e906884ccf35decf29b5499e51412d --- /dev/null +++ b/outputs/2026-06-20/12-58-54/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-20/12-58-54/main.log b/outputs/2026-06-20/12-58-54/main.log new file mode 100644 index 0000000000000000000000000000000000000000..bc38b646c4fcb1a1f4aac0b573033d145c523359 --- /dev/null +++ b/outputs/2026-06-20/12-58-54/main.log @@ -0,0 +1,3 @@ +[2026-06-20 12:59:40,037][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-20 13:00:04,596][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 13:00:07,035][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-20/12-58-55/.hydra/config.yaml b/outputs/2026-06-20/12-58-55/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..faf4fdc2cdedc288bdd5309e6df8362385b2f4ea --- /dev/null +++ b/outputs/2026-06-20/12-58-55/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: all + dynamic: all + revisit: all + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-20/12-58-55/.hydra/hydra.yaml b/outputs/2026-06-20/12-58-55/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e37a4fc2120f4967d212e9a378998c194b0ed511 --- /dev/null +++ b/outputs/2026-06-20/12-58-55/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=all + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=all + - ++algorithm.dememwm.frame_memory.noise_route.revisit=all + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=0,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=all,++algorithm.dememwm.frame_memory.noise_route.dynamic=all,++algorithm.dememwm.frame_memory.noise_route.revisit=all,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-20/12-58-55 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-20/12-58-55/.hydra/overrides.yaml b/outputs/2026-06-20/12-58-55/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9ea76abe63e906884ccf35decf29b5499e51412d --- /dev/null +++ b/outputs/2026-06-20/12-58-55/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_no_noise_route_no_dynamic_a2r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=0 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=all +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=all +- ++algorithm.dememwm.frame_memory.noise_route.revisit=all +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-20/12-58-55/main.log b/outputs/2026-06-20/12-58-55/main.log new file mode 100644 index 0000000000000000000000000000000000000000..312bc16f3e67b045f4c3c2ab47992e8cde200eaf --- /dev/null +++ b/outputs/2026-06-20/12-58-55/main.log @@ -0,0 +1,3 @@ +[2026-06-20 12:59:40,037][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-20 13:00:04,596][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 13:00:07,035][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-20/13-04-59/.hydra/config.yaml b/outputs/2026-06-20/13-04-59/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4981e7b728f86c287b16ee6a10121000a82e701e --- /dev/null +++ b/outputs/2026-06-20/13-04-59/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: low + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-20/13-04-59/.hydra/hydra.yaml b/outputs/2026-06-20/13-04-59/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..48c711d5fe20dbdd2d55c3e8393b4631f8a77644 --- /dev/null +++ b/outputs/2026-06-20/13-04-59/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=low + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=low,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-20/13-04-59 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-20/13-04-59/.hydra/overrides.yaml b/outputs/2026-06-20/13-04-59/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..07967ff3a9545c962a89eaa7556655ddd496dad9 --- /dev/null +++ b/outputs/2026-06-20/13-04-59/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=low +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-20/13-04-59/main.log b/outputs/2026-06-20/13-04-59/main.log new file mode 100644 index 0000000000000000000000000000000000000000..344b4ab2839779fd8b92c3935cf0247a2a9921c5 --- /dev/null +++ b/outputs/2026-06-20/13-04-59/main.log @@ -0,0 +1,40 @@ +[2026-06-20 13:05:47,204][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-20 13:05:47,479][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-20 13:05:47,479][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-20 13:05:47,479][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-20 13:05:47,479][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-20 13:05:47,479][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-20 13:05:49,903][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: matched=302 missing=0 shape_mismatch=0 prefix='' +[2026-06-20 13:05:51,092][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: matched=228 missing=0 shape_mismatch=0 prefix='' +[2026-06-20 13:05:53,479][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-20 13:05:53,479][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-20 13:05:53,479][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-20 13:05:53,484][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-20 13:05:53,500][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-20 13:05:58,974][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-20 13:06:18,966][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 13:06:18,966][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 13:06:18,966][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 13:06:18,966][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 13:06:18,966][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 13:06:19,956][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +1.0 B Trainable params +231 M Non-trainable params +1.3 B Total params +5,107.144 Total estimated model params size (MB) +[2026-06-20 13:06:22,589][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-20 13:06:22,589][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-20 13:06:22,589][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-20 13:06:22,589][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-20 13:06:22,589][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-22 21:41:21,456][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=175000` reached. diff --git a/outputs/2026-06-20/13-05-00/.hydra/config.yaml b/outputs/2026-06-20/13-05-00/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4981e7b728f86c287b16ee6a10121000a82e701e --- /dev/null +++ b/outputs/2026-06-20/13-05-00/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: low + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-20/13-05-00/.hydra/hydra.yaml b/outputs/2026-06-20/13-05-00/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c4f5f2088d9b629296440f3c226f00a1b8dbc897 --- /dev/null +++ b/outputs/2026-06-20/13-05-00/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=low + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=low,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-20/13-05-00 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-20/13-05-00/.hydra/overrides.yaml b/outputs/2026-06-20/13-05-00/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..07967ff3a9545c962a89eaa7556655ddd496dad9 --- /dev/null +++ b/outputs/2026-06-20/13-05-00/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=low +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-20/13-05-00/main.log b/outputs/2026-06-20/13-05-00/main.log new file mode 100644 index 0000000000000000000000000000000000000000..5d9fe1fe0f2bd836b1c7b88a1cdfc0c4a95945ba --- /dev/null +++ b/outputs/2026-06-20/13-05-00/main.log @@ -0,0 +1,6 @@ +[2026-06-20 13:05:53,479][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-20 13:05:53,479][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-20 13:06:18,966][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 13:06:18,966][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 13:06:22,589][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-20 13:06:22,589][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-20/13-05-01/.hydra/config.yaml b/outputs/2026-06-20/13-05-01/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4981e7b728f86c287b16ee6a10121000a82e701e --- /dev/null +++ b/outputs/2026-06-20/13-05-01/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 175000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + save_last: true + enable_version_counter: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: true + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + metadata: data/${dataset.name}/metadata.json + data_mean: 0.5 + data_std: 0.5 + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + n_frames: 1000 + context_length: 1 + resolution: 128 + observation_shape: + - 3 + - ${dataset.resolution} + - ${dataset.resolution} + external_cond_dim: 0 + validation_multiplier: 1 + frame_skip: 1 + action_cond_dim: 25 + memory_condition_length: 0 + customized_validation: true + angle_range: 180 + pos_range: 8 + max_getitem_retries: 32 + _name: video_minecraft_latent + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + n_frames_valid: 500 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: + - 16 + - 18 + - 32 + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 100 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + action_cond_dim: 25 + use_plucker: true + _name: dememwm_frame_memory_ablation + memory_token_cross_attention: false + pose_cond_dim: 5 + log_video: true + dememwm: + enabled: true + debug_stdout_metrics: true + training_stage: stage_2 + curriculum: + enabled: true + full_stage_start_step: 0 + freeze_vae: true + lr: + dememwm_modules: 8.0e-05 + memory_adapters: 8.0e-05 + full_dit: 2.0e-05 + frame_memory: + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + dynamic_temporal: bidirectional_stream + anchor_temporal: diagonal + revisit_temporal: diagonal + noise_route: + anchor: high + dynamic: low + revisit: low + anchor_diverse_selection: true + chunk_retrieval: + chunk_fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + allow_empty_revisit: true + candidate_chunk_size: 64 + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + revisit_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit: + deterministic_pose_retrieval: true + fov_half_h: 52.5 + fov_half_v: 37.5 + fov_radius: 30.0 + fov_yaw_samples: 25 + fov_pitch_samples: 20 + fov_depth_samples: 20 + pose_preselect_topk: 64 + plucker_weight: 0.1 + plucker_grid_h: 4 + plucker_grid_w: 4 + plucker_focal_length: 0.35 + revisit_metric: + enabled: false + min_gap: 8 + gap_bands: + - 32 + - 128 + fov_overlap_threshold: 0.6 + positional_distance_threshold: 2.0 + self_crop_fraction: 0.5 + cache: + enabled: false + checkpoint: + strict_dememwm_eval_load: true + n_tokens: 8 +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +only_tune_memory: false diff --git a/outputs/2026-06-20/13-05-01/.hydra/hydra.yaml b/outputs/2026-06-20/13-05-01/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..dd7bf506276880e476bb7790deb98c60f70876d5 --- /dev/null +++ b/outputs/2026-06-20/13-05-01/.hydra/hydra.yaml @@ -0,0 +1,217 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k/ + - auto_resume=true + - experiment.tasks=[training] + - algorithm=dememwm_frame_memory_ablation + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - +only_tune_memory=false + - dataset=video_minecraft_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=1000 + - +dataset.n_frames_valid=500 + - dataset.customized_validation=true + - +dataset.wo_updown=false + - dataset.angle_range=180 + - dataset.pos_range=8 + - ++algorithm.n_tokens=8 + - algorithm.x_shape=[16,18,32] + - ++algorithm.context_frames=100 + - ++algorithm.log_video=true + - ++algorithm.diffusion.sampling_timesteps=20 + - ++algorithm.dememwm.debug_stdout_metrics=true + - ++algorithm.dememwm.revisit_metric.enabled=false + - ++algorithm.dememwm.frame_memory.max_anchor_frames=2 + - ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 + - ++algorithm.dememwm.frame_memory.max_revisit_frames=2 + - ++algorithm.dememwm.frame_memory.noise_route.anchor=high + - ++algorithm.dememwm.frame_memory.noise_route.dynamic=low + - ++algorithm.dememwm.frame_memory.noise_route.revisit=low + - ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true + - ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 + - ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 + - ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true + - ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 + - ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 + - ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 + - ++algorithm.dememwm.revisit.fov_yaw_samples=25 + - ++algorithm.dememwm.revisit.fov_pitch_samples=20 + - ++algorithm.dememwm.revisit.fov_depth_samples=20 + - ++algorithm.dememwm.revisit.plucker_weight=0.10 + - ++algorithm.dememwm.revisit.plucker_focal_length=0.35 + - ++algorithm.dememwm.revisit.pose_preselect_topk=64 + - ++algorithm.dememwm.curriculum.enabled=true + - ++algorithm.dememwm.curriculum.full_stage_start_step=0 + - ++algorithm.dememwm.curriculum.freeze_vae=true + - ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 + - ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=175000 + job: + name: main + chdir: null + override_dirname: ++algorithm.context_frames=100,++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true,++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64,++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60,++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10,++algorithm.dememwm.curriculum.enabled=true,++algorithm.dememwm.curriculum.freeze_vae=true,++algorithm.dememwm.curriculum.full_stage_start_step=0,++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5,++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5,++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5,++algorithm.dememwm.debug_stdout_metrics=true,++algorithm.dememwm.frame_memory.anchor_diverse_selection=true,++algorithm.dememwm.frame_memory.max_anchor_frames=2,++algorithm.dememwm.frame_memory.max_dynamic_frames=4,++algorithm.dememwm.frame_memory.max_revisit_frames=2,++algorithm.dememwm.frame_memory.noise_route.anchor=high,++algorithm.dememwm.frame_memory.noise_route.dynamic=low,++algorithm.dememwm.frame_memory.noise_route.revisit=low,++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25,++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50,++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25,++algorithm.dememwm.revisit.fov_depth_samples=20,++algorithm.dememwm.revisit.fov_pitch_samples=20,++algorithm.dememwm.revisit.fov_yaw_samples=25,++algorithm.dememwm.revisit.plucker_focal_length=0.35,++algorithm.dememwm.revisit.plucker_weight=0.10,++algorithm.dememwm.revisit.pose_preselect_topk=64,++algorithm.dememwm.revisit_metric.enabled=false,++algorithm.diffusion.sampling_timesteps=20,++algorithm.log_video=true,++algorithm.n_tokens=8,+customized_load=true,+dataset.n_frames_valid=500,+dataset.wo_updown=false,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k,+only_tune_memory=false,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.x_shape=[16,18,32],algorithm=dememwm_frame_memory_ablation,auto_resume=true,dataset.angle_range=180,dataset.customized_validation=true,dataset.n_frames=1000,dataset.pos_range=8,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset=video_minecraft_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.max_steps=175000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000 + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-20/13-05-01 + choices: + cluster: null + algorithm: dememwm_frame_memory_ablation + dataset: video_minecraft_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-20/13-05-01/.hydra/overrides.yaml b/outputs/2026-06-20/13-05-01/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..07967ff3a9545c962a89eaa7556655ddd496dad9 --- /dev/null +++ b/outputs/2026-06-20/13-05-01/.hydra/overrides.yaml @@ -0,0 +1,59 @@ +- +name=train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sep_mem_fulldit_mem_route_anchor_high_dynamic_revisit_low_a2d4r2_ntok8_8h200_bs16_global128_175k/ +- auto_resume=true +- experiment.tasks=[training] +- algorithm=dememwm_frame_memory_ablation +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- +only_tune_memory=false +- dataset=video_minecraft_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=1000 +- +dataset.n_frames_valid=500 +- dataset.customized_validation=true +- +dataset.wo_updown=false +- dataset.angle_range=180 +- dataset.pos_range=8 +- ++algorithm.n_tokens=8 +- algorithm.x_shape=[16,18,32] +- ++algorithm.context_frames=100 +- ++algorithm.log_video=true +- ++algorithm.diffusion.sampling_timesteps=20 +- ++algorithm.dememwm.debug_stdout_metrics=true +- ++algorithm.dememwm.revisit_metric.enabled=false +- ++algorithm.dememwm.frame_memory.max_anchor_frames=2 +- ++algorithm.dememwm.frame_memory.max_dynamic_frames=4 +- ++algorithm.dememwm.frame_memory.max_revisit_frames=2 +- ++algorithm.dememwm.frame_memory.noise_route.anchor=high +- ++algorithm.dememwm.frame_memory.noise_route.dynamic=low +- ++algorithm.dememwm.frame_memory.noise_route.revisit=low +- ++algorithm.dememwm.frame_memory.anchor_diverse_selection=true +- ++algorithm.dememwm.chunk_retrieval.chunk_fov_overlap_threshold=0.60 +- ++algorithm.dememwm.chunk_retrieval.min_total_selected_coverage=0.10 +- ++algorithm.dememwm.chunk_retrieval.allow_empty_revisit=true +- ++algorithm.dememwm.chunk_retrieval.candidate_chunk_size=64 +- ++algorithm.dememwm.memory_noise.anchor_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.revisit_max_fraction=0.25 +- ++algorithm.dememwm.memory_noise.dynamic_max_fraction=0.50 +- ++algorithm.dememwm.revisit.fov_yaw_samples=25 +- ++algorithm.dememwm.revisit.fov_pitch_samples=20 +- ++algorithm.dememwm.revisit.fov_depth_samples=20 +- ++algorithm.dememwm.revisit.plucker_weight=0.10 +- ++algorithm.dememwm.revisit.plucker_focal_length=0.35 +- ++algorithm.dememwm.revisit.pose_preselect_topk=64 +- ++algorithm.dememwm.curriculum.enabled=true +- ++algorithm.dememwm.curriculum.full_stage_start_step=0 +- ++algorithm.dememwm.curriculum.freeze_vae=true +- ++algorithm.dememwm.curriculum.lr.dememwm_modules=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.memory_adapters=8.0e-5 +- ++algorithm.dememwm.curriculum.lr.full_dit=2.0e-5 +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=175000 diff --git a/outputs/2026-06-20/13-05-01/main.log b/outputs/2026-06-20/13-05-01/main.log new file mode 100644 index 0000000000000000000000000000000000000000..a347d9cb474fc066c6039ad6a9bdec7762dfa054 --- /dev/null +++ b/outputs/2026-06-20/13-05-01/main.log @@ -0,0 +1,3 @@ +[2026-06-20 13:05:53,479][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-20 13:06:18,966][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-20 13:06:22,589][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-25/20-41-58/.hydra/config.yaml b/outputs/2026-06-25/20-41-58/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7c97550b32338ca18b91468ee081745e758dd303 --- /dev/null +++ b/outputs/2026-06-25/20-41-58/.hydra/config.yaml @@ -0,0 +1,211 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 60000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 700 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: recent + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 5000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 20000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-25/20-41-58/.hydra/hydra.yaml b/outputs/2026-06-25/20-41-58/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..68b652c16c00981e19a3056a0b6248bb532f7cc4 --- /dev/null +++ b/outputs/2026-06-25/20-41-58/.hydra/hydra.yaml @@ -0,0 +1,216 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=700 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=recent + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=5000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=20000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_max_fraction=0.50 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_max_fraction=0.50,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=20000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=5000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.dynamic.selection_policy=recent,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=700,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-25/20-41-58 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-25/20-41-58/.hydra/overrides.yaml b/outputs/2026-06-25/20-41-58/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..634c9c262ff8c9a6fac58a418e60d4661df1695a --- /dev/null +++ b/outputs/2026-06-25/20-41-58/.hydra/overrides.yaml @@ -0,0 +1,58 @@ +- +name=train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=700 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=recent +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=5000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=20000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_max_fraction=0.50 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-06-25/20-41-58/main.log b/outputs/2026-06-25/20-41-58/main.log new file mode 100644 index 0000000000000000000000000000000000000000..8a464c9bda4df31b449317d213b4baad503a4d7c --- /dev/null +++ b/outputs/2026-06-25/20-41-58/main.log @@ -0,0 +1,87 @@ +[2026-06-25 20:42:51,361][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-06-25 20:42:51,362][pytorch_lightning.utilities.rank_zero][INFO] - Missing checkpoint keys: timestamp_embedding.mlp.0.weight, timestamp_embedding.mlp.0.bias, timestamp_embedding.mlp.2.weight, timestamp_embedding.mlp.2.bias, spatial_rotary_emb.freqs, temporal_rotary_emb.freqs, pose_embedder.model.0.weight, pose_embedder.model.0.bias, ... (+560 more) +[2026-06-25 20:42:53,680][pytorch_lightning.utilities.rank_zero][INFO] - Zeroed frame-memory gate slices for base checkpoint load: blocks.0.r_adaLN_modulation.1.weight, blocks.0.r_adaLN_modulation.1.bias, blocks.1.r_adaLN_modulation.1.weight, blocks.1.r_adaLN_modulation.1.bias, blocks.2.r_adaLN_modulation.1.weight, blocks.2.r_adaLN_modulation.1.bias, blocks.3.r_adaLN_modulation.1.weight, blocks.3.r_adaLN_modulation.1.bias, ... (+24 more) +[2026-06-25 20:42:53,680][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: 302 tensors. +[2026-06-25 20:42:53,923][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-06-25 20:42:54,998][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: 228 tensors. +[2026-06-25 20:42:56,277][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-25 20:42:56,419][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-25 20:42:56,419][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-25 20:42:56,419][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-25 20:42:56,419][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-25 20:42:56,419][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-25 20:43:50,238][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-25 20:43:50,239][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-25 20:43:50,758][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-25 20:43:57,072][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-25 20:44:16,384][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:44:16,384][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:44:16,384][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:44:18,116][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-25 20:44:20,911][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:44:20,912][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:44:20,912][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 22:04:28,124][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-25 22:04:28,866][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-25 22:04:28,867][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-25 22:04:28,867][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-26 13:48:08,874][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=60000` reached. +[2026-06-26 13:48:40,202][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-26 13:48:40,203][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:48:40,203][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:48:40,203][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:48:40,984][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-26 13:48:40,984][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-26 13:48:40,984][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-26 13:48:40,984][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-26 13:48:40,985][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-26 13:50:40,309][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/callbacks/model_checkpoint.py:639: Checkpoint directory /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints exists and is not empty. + +[2026-06-26 13:50:40,310][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 13:50:40,329][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:50:40,331][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:50:40,334][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:51:01,538][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:51:01,648][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:51:01,656][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:51:02,797][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-26 13:51:03,960][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 13:51:04,134][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:51:04,134][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:51:04,135][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:51:06,389][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/loops/training_epoch_loop.py:154: You're resuming from a checkpoint that ended before the epoch ended. This can cause unreliable results if further training is done. Consider using an end-of-epoch checkpoint + diff --git a/outputs/2026-06-25/20-41-59/.hydra/config.yaml b/outputs/2026-06-25/20-41-59/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7c97550b32338ca18b91468ee081745e758dd303 --- /dev/null +++ b/outputs/2026-06-25/20-41-59/.hydra/config.yaml @@ -0,0 +1,211 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 60000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 700 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: recent + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 5000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 20000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-25/20-41-59/.hydra/hydra.yaml b/outputs/2026-06-25/20-41-59/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4437b5cdc8ee17d2a3d0579efc896041bc309b2b --- /dev/null +++ b/outputs/2026-06-25/20-41-59/.hydra/hydra.yaml @@ -0,0 +1,216 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=700 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=recent + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=5000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=20000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_max_fraction=0.50 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_max_fraction=0.50,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=20000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=5000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.dynamic.selection_policy=recent,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=700,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-25/20-41-59 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-25/20-41-59/.hydra/overrides.yaml b/outputs/2026-06-25/20-41-59/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..634c9c262ff8c9a6fac58a418e60d4661df1695a --- /dev/null +++ b/outputs/2026-06-25/20-41-59/.hydra/overrides.yaml @@ -0,0 +1,58 @@ +- +name=train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=700 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=recent +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=5000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=20000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_max_fraction=0.50 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-06-25/20-41-59/main.log b/outputs/2026-06-25/20-41-59/main.log new file mode 100644 index 0000000000000000000000000000000000000000..c35b0fda4f4b1abe53640231fed351fc88414209 --- /dev/null +++ b/outputs/2026-06-25/20-41-59/main.log @@ -0,0 +1,45 @@ +[2026-06-25 20:43:50,238][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-25 20:43:50,241][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-25 20:43:50,251][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-25 20:43:50,260][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-25 20:43:51,094][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-25 20:44:16,384][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:44:16,384][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:44:16,384][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:44:16,384][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:44:16,384][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:44:20,911][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:44:20,912][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:44:20,912][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:44:20,912][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:44:20,912][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:48:40,203][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:48:40,203][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:48:40,203][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:48:40,202][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:48:40,203][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:50:40,317][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:50:40,323][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:50:40,325][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:50:40,327][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:50:40,336][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:51:01,558][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:51:01,573][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:51:01,662][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:51:01,664][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:51:01,750][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:51:04,134][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:51:04,134][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:51:04,134][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:51:04,134][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:51:04,135][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-25/20-46-55/.hydra/config.yaml b/outputs/2026-06-25/20-46-55/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7021a9adaba2f265eb735cd2e1ba34e1b0e98274 --- /dev/null +++ b/outputs/2026-06-25/20-46-55/.hydra/config.yaml @@ -0,0 +1,211 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 60000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 700 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 5000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 20000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-25/20-46-55/.hydra/hydra.yaml b/outputs/2026-06-25/20-46-55/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3257555fd7dec6955945b2ca5d1a36b84cbd01e2 --- /dev/null +++ b/outputs/2026-06-25/20-46-55/.hydra/hydra.yaml @@ -0,0 +1,216 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=700 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=5000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=20000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_max_fraction=0.50 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_max_fraction=0.50,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=20000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=5000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=700,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-25/20-46-55 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-25/20-46-55/.hydra/overrides.yaml b/outputs/2026-06-25/20-46-55/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..50839acd41803ee9a5606af6358b422484826767 --- /dev/null +++ b/outputs/2026-06-25/20-46-55/.hydra/overrides.yaml @@ -0,0 +1,58 @@ +- +name=train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=700 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=5000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=20000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_max_fraction=0.50 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-06-25/20-46-55/main.log b/outputs/2026-06-25/20-46-55/main.log new file mode 100644 index 0000000000000000000000000000000000000000..48cc19ef311ff80b70541d2c34db07b18b07ad6c --- /dev/null +++ b/outputs/2026-06-25/20-46-55/main.log @@ -0,0 +1,18 @@ +[2026-06-25 20:48:35,989][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-25 20:48:36,069][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-25 20:49:00,760][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:49:00,760][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:49:04,366][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:49:04,366][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:55:02,317][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:55:02,319][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:56:50,190][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:56:50,193][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:57:10,231][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:57:10,343][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:57:12,313][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:57:12,313][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-25/20-46-56/.hydra/config.yaml b/outputs/2026-06-25/20-46-56/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7021a9adaba2f265eb735cd2e1ba34e1b0e98274 --- /dev/null +++ b/outputs/2026-06-25/20-46-56/.hydra/config.yaml @@ -0,0 +1,211 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 60000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 700 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 5000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 20000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-25/20-46-56/.hydra/hydra.yaml b/outputs/2026-06-25/20-46-56/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..947816c6fc5eca2c3fcd6955be6425ada8871ca2 --- /dev/null +++ b/outputs/2026-06-25/20-46-56/.hydra/hydra.yaml @@ -0,0 +1,216 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=700 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=5000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=20000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_max_fraction=0.50 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_max_fraction=0.50,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=20000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=5000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=700,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-25/20-46-56 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-25/20-46-56/.hydra/overrides.yaml b/outputs/2026-06-25/20-46-56/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..50839acd41803ee9a5606af6358b422484826767 --- /dev/null +++ b/outputs/2026-06-25/20-46-56/.hydra/overrides.yaml @@ -0,0 +1,58 @@ +- +name=train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=700 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=5000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=20000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_max_fraction=0.50 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-06-25/20-46-56/main.log b/outputs/2026-06-25/20-46-56/main.log new file mode 100644 index 0000000000000000000000000000000000000000..b3091bb38a0087a912cc65d3af471e9ed1baedb3 --- /dev/null +++ b/outputs/2026-06-25/20-46-56/main.log @@ -0,0 +1,114 @@ +[2026-06-25 20:47:42,438][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-06-25 20:47:42,439][pytorch_lightning.utilities.rank_zero][INFO] - Missing checkpoint keys: timestamp_embedding.mlp.0.weight, timestamp_embedding.mlp.0.bias, timestamp_embedding.mlp.2.weight, timestamp_embedding.mlp.2.bias, spatial_rotary_emb.freqs, temporal_rotary_emb.freqs, pose_embedder.model.0.weight, pose_embedder.model.0.bias, ... (+560 more) +[2026-06-25 20:47:44,318][pytorch_lightning.utilities.rank_zero][INFO] - Zeroed frame-memory gate slices for base checkpoint load: blocks.0.r_adaLN_modulation.1.weight, blocks.0.r_adaLN_modulation.1.bias, blocks.1.r_adaLN_modulation.1.weight, blocks.1.r_adaLN_modulation.1.bias, blocks.2.r_adaLN_modulation.1.weight, blocks.2.r_adaLN_modulation.1.bias, blocks.3.r_adaLN_modulation.1.weight, blocks.3.r_adaLN_modulation.1.bias, ... (+24 more) +[2026-06-25 20:47:44,319][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: 302 tensors. +[2026-06-25 20:47:44,523][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-06-25 20:47:45,539][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: 228 tensors. +[2026-06-25 20:47:47,472][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-25 20:47:47,621][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-25 20:47:47,621][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-25 20:47:47,621][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-25 20:47:47,621][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-25 20:47:47,621][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-25 20:48:35,983][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-25 20:48:35,983][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-25 20:48:35,984][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-25 20:48:35,984][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-25 20:48:35,985][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-25 20:48:35,987][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-25 20:48:41,626][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-25 20:49:00,760][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:49:00,760][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:49:00,760][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:49:00,760][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:49:00,760][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:49:00,760][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:49:03,680][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-25 20:49:04,366][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:49:04,366][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:49:04,366][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:49:04,366][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:49:04,366][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:49:04,366][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 22:09:23,939][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-25 22:09:25,688][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-25 22:09:25,688][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-25 22:09:25,688][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-26 13:54:35,992][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=60000` reached. +[2026-06-26 13:55:02,317][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-26 13:55:02,317][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:55:02,317][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:55:02,317][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:55:02,318][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:55:02,319][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:55:02,319][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:55:03,826][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-26 13:55:03,826][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-26 13:55:03,826][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-26 13:55:03,826][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-26 13:55:03,827][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-26 13:56:50,180][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/callbacks/model_checkpoint.py:639: Checkpoint directory /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints exists and is not empty. + +[2026-06-26 13:56:50,181][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 13:56:50,188][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:56:50,195][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:56:50,197][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:56:50,199][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:56:50,201][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:56:50,202][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:57:10,226][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:57:10,232][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:57:10,234][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:57:10,237][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:57:10,238][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:57:10,241][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:57:11,151][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-26 13:57:12,280][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 13:57:12,313][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:57:12,313][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:57:12,313][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:57:12,313][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:57:12,313][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:57:12,313][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:57:13,295][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/loops/training_epoch_loop.py:154: You're resuming from a checkpoint that ended before the epoch ended. This can cause unreliable results if further training is done. Consider using an end-of-epoch checkpoint + diff --git a/outputs/2026-06-25/20-48-27/.hydra/config.yaml b/outputs/2026-06-25/20-48-27/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3e0d1eacb4a756bcfda7d47203d6772fdda4254a --- /dev/null +++ b/outputs/2026-06-25/20-48-27/.hydra/config.yaml @@ -0,0 +1,211 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 60000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 700 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 5000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: false + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 20000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-25/20-48-27/.hydra/hydra.yaml b/outputs/2026-06-25/20-48-27/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..04f338bb45c36606bcf0ad6ac57403461f5cc2eb --- /dev/null +++ b/outputs/2026-06-25/20-48-27/.hydra/hydra.yaml @@ -0,0 +1,216 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=700 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=5000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=20000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=false + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_max_fraction=0.50 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_max_fraction=0.50,algorithm.memory_noise.enabled=false,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=20000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=5000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=700,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-25/20-48-27 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-25/20-48-27/.hydra/overrides.yaml b/outputs/2026-06-25/20-48-27/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5c8056a58520299ba8204bad4bbcc640e272ce00 --- /dev/null +++ b/outputs/2026-06-25/20-48-27/.hydra/overrides.yaml @@ -0,0 +1,58 @@ +- +name=train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=700 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=5000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=20000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=false +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_max_fraction=0.50 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-06-25/20-48-27/main.log b/outputs/2026-06-25/20-48-27/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e3243c1a33b535af1489abd14b9e3877714b2c8f --- /dev/null +++ b/outputs/2026-06-25/20-48-27/main.log @@ -0,0 +1,123 @@ +[2026-06-25 20:49:16,230][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-06-25 20:49:16,231][pytorch_lightning.utilities.rank_zero][INFO] - Missing checkpoint keys: timestamp_embedding.mlp.0.weight, timestamp_embedding.mlp.0.bias, timestamp_embedding.mlp.2.weight, timestamp_embedding.mlp.2.bias, spatial_rotary_emb.freqs, temporal_rotary_emb.freqs, pose_embedder.model.0.weight, pose_embedder.model.0.bias, ... (+560 more) +[2026-06-25 20:49:18,375][pytorch_lightning.utilities.rank_zero][INFO] - Zeroed frame-memory gate slices for base checkpoint load: blocks.0.r_adaLN_modulation.1.weight, blocks.0.r_adaLN_modulation.1.bias, blocks.1.r_adaLN_modulation.1.weight, blocks.1.r_adaLN_modulation.1.bias, blocks.2.r_adaLN_modulation.1.weight, blocks.2.r_adaLN_modulation.1.bias, blocks.3.r_adaLN_modulation.1.weight, blocks.3.r_adaLN_modulation.1.bias, ... (+24 more) +[2026-06-25 20:49:18,375][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: 302 tensors. +[2026-06-25 20:49:18,540][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-06-25 20:49:19,633][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: 228 tensors. +[2026-06-25 20:49:20,614][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-25 20:49:20,769][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-25 20:49:20,769][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-25 20:49:20,769][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-25 20:49:20,769][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-25 20:49:20,770][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-25 20:50:10,880][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-25 20:50:10,880][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-25 20:50:10,881][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-25 20:50:10,897][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-25 20:50:10,913][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-25 20:50:10,922][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-25 20:50:11,941][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-25 20:50:18,975][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-25 20:50:37,831][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:50:37,831][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:50:37,831][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:50:37,831][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:50:37,831][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:50:37,831][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:50:37,831][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:50:39,913][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-25 20:50:39,915][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:50:39,915][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:50:39,915][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:50:39,915][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:50:39,915][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:50:39,915][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 20:50:39,915][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 22:10:55,981][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-25 22:11:02,035][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-25 22:11:02,035][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-25 22:11:02,036][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-26 13:54:55,308][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=60000` reached. +[2026-06-26 13:55:15,902][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-26 13:55:15,902][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:55:15,902][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:55:15,902][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:55:15,902][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:55:15,902][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:55:15,902][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:55:15,902][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:55:16,851][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-26 13:55:16,851][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-26 13:55:16,851][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-26 13:55:16,851][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-26 13:55:16,852][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-26 13:57:08,336][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/callbacks/model_checkpoint.py:639: Checkpoint directory /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints exists and is not empty. + +[2026-06-26 13:57:08,337][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 13:57:08,347][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:57:08,349][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:57:08,351][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:57:08,353][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:57:08,355][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:57:08,356][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:57:08,358][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:57:27,190][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:57:27,192][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:57:27,195][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:57:27,198][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:57:27,199][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:57:27,204][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:57:27,307][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:57:28,097][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-26 13:57:29,255][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 13:57:29,263][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:57:29,263][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:57:29,263][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:57:29,263][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:57:29,263][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:57:29,263][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:57:29,263][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:57:30,277][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/loops/training_epoch_loop.py:154: You're resuming from a checkpoint that ended before the epoch ended. This can cause unreliable results if further training is done. Consider using an end-of-epoch checkpoint + diff --git a/outputs/2026-06-25/20-48-28/.hydra/config.yaml b/outputs/2026-06-25/20-48-28/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3e0d1eacb4a756bcfda7d47203d6772fdda4254a --- /dev/null +++ b/outputs/2026-06-25/20-48-28/.hydra/config.yaml @@ -0,0 +1,211 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 60000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 700 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 5000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: false + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 20000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-25/20-48-28/.hydra/hydra.yaml b/outputs/2026-06-25/20-48-28/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5f79c7d1c86b842405a68921f03971491b3bc969 --- /dev/null +++ b/outputs/2026-06-25/20-48-28/.hydra/hydra.yaml @@ -0,0 +1,216 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=700 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=5000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=20000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=false + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_max_fraction=0.50 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_max_fraction=0.50,algorithm.memory_noise.enabled=false,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=20000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=5000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=700,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-25/20-48-28 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-25/20-48-28/.hydra/overrides.yaml b/outputs/2026-06-25/20-48-28/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5c8056a58520299ba8204bad4bbcc640e272ce00 --- /dev/null +++ b/outputs/2026-06-25/20-48-28/.hydra/overrides.yaml @@ -0,0 +1,58 @@ +- +name=train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=700 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=5000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=20000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=false +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_max_fraction=0.50 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-06-25/20-48-28/main.log b/outputs/2026-06-25/20-48-28/main.log new file mode 100644 index 0000000000000000000000000000000000000000..d0f354cdeaa2caeb90386f108b9a8528bd1e7da2 --- /dev/null +++ b/outputs/2026-06-25/20-48-28/main.log @@ -0,0 +1,9 @@ +[2026-06-25 20:50:10,882][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-25 20:50:37,831][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 20:50:39,915][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 13:55:15,902][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 13:57:08,342][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 13:57:27,198][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 13:57:29,263][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-25/21-28-59/.hydra/config.yaml b/outputs/2026-06-25/21-28-59/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d24738b99d91e3ae14cbbe8a4c211bcf4b2fc76b --- /dev/null +++ b/outputs/2026-06-25/21-28-59/.hydra/config.yaml @@ -0,0 +1,211 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 60000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 700 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 5000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + noise_route: + anchor: all + dynamic: all + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 20000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-25/21-28-59/.hydra/hydra.yaml b/outputs/2026-06-25/21-28-59/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ed19374bfb472c12c6ce2d904c84bd2fe664bd0e --- /dev/null +++ b/outputs/2026-06-25/21-28-59/.hydra/hydra.yaml @@ -0,0 +1,216 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=700 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=5000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=20000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=all + - algorithm.noise_route.dynamic=all + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_max_fraction=0.50 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_max_fraction=0.50,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.n_tokens=8,algorithm.noise_route.anchor=all,algorithm.noise_route.dynamic=all,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=20000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=5000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=700,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-25/21-28-59 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-25/21-28-59/.hydra/overrides.yaml b/outputs/2026-06-25/21-28-59/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4b85025abd8fab687599f0ac1ed0ad4c8ddb9f9a --- /dev/null +++ b/outputs/2026-06-25/21-28-59/.hydra/overrides.yaml @@ -0,0 +1,58 @@ +- +name=train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=700 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=5000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=20000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=all +- algorithm.noise_route.dynamic=all +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_max_fraction=0.50 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-06-25/21-28-59/main.log b/outputs/2026-06-25/21-28-59/main.log new file mode 100644 index 0000000000000000000000000000000000000000..924f3eab48ffbf6024f644a2a64e88d678d0e23b --- /dev/null +++ b/outputs/2026-06-25/21-28-59/main.log @@ -0,0 +1,123 @@ +[2026-06-25 21:29:44,339][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-06-25 21:29:44,339][pytorch_lightning.utilities.rank_zero][INFO] - Missing checkpoint keys: timestamp_embedding.mlp.0.weight, timestamp_embedding.mlp.0.bias, timestamp_embedding.mlp.2.weight, timestamp_embedding.mlp.2.bias, spatial_rotary_emb.freqs, temporal_rotary_emb.freqs, pose_embedder.model.0.weight, pose_embedder.model.0.bias, ... (+560 more) +[2026-06-25 21:29:44,904][pytorch_lightning.utilities.rank_zero][INFO] - Zeroed frame-memory gate slices for base checkpoint load: blocks.0.r_adaLN_modulation.1.weight, blocks.0.r_adaLN_modulation.1.bias, blocks.1.r_adaLN_modulation.1.weight, blocks.1.r_adaLN_modulation.1.bias, blocks.2.r_adaLN_modulation.1.weight, blocks.2.r_adaLN_modulation.1.bias, blocks.3.r_adaLN_modulation.1.weight, blocks.3.r_adaLN_modulation.1.bias, ... (+24 more) +[2026-06-25 21:29:44,905][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: 302 tensors. +[2026-06-25 21:29:45,157][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-06-25 21:29:46,389][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: 228 tensors. +[2026-06-25 21:29:47,572][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-25 21:29:47,704][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-25 21:29:47,704][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-25 21:29:47,704][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-25 21:29:47,704][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-25 21:29:47,705][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-25 21:30:32,154][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-25 21:30:32,154][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-25 21:30:32,154][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-25 21:30:32,154][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-25 21:30:32,155][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-25 21:30:32,174][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-25 21:30:32,439][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-25 21:30:38,009][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-25 21:30:56,337][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 21:30:56,337][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 21:30:56,337][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 21:30:56,337][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 21:30:56,337][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 21:30:56,337][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 21:30:56,337][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 21:30:57,269][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-25 21:30:58,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 21:30:58,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 21:30:58,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 21:30:58,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 21:30:58,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 21:30:58,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 21:30:58,776][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-25 22:51:12,834][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-25 22:51:17,265][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-25 22:51:17,265][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-25 22:51:17,265][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-26 14:36:39,200][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=60000` reached. +[2026-06-26 14:37:00,532][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 14:37:00,533][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-26 14:37:00,533][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 14:37:00,533][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 14:37:00,533][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 14:37:00,533][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 14:37:00,534][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 14:37:00,534][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 14:37:00,869][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-26 14:37:00,869][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-26 14:37:00,869][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-26 14:37:00,869][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-26 14:37:00,870][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-26 14:38:39,385][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/callbacks/model_checkpoint.py:639: Checkpoint directory /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/checkpoints exists and is not empty. + +[2026-06-26 14:38:39,386][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 14:38:39,390][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 14:38:39,395][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 14:38:39,397][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 14:38:39,398][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 14:38:39,400][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 14:38:39,402][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 14:38:39,404][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 14:39:01,288][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 14:39:01,292][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 14:39:01,293][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 14:39:01,293][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 14:39:01,310][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 14:39:01,310][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 14:39:01,317][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 14:39:02,200][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-26 14:39:03,357][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 14:39:03,399][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 14:39:03,399][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 14:39:03,399][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 14:39:03,399][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 14:39:03,399][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 14:39:03,399][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 14:39:03,399][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 14:39:05,104][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/loops/training_epoch_loop.py:154: You're resuming from a checkpoint that ended before the epoch ended. This can cause unreliable results if further training is done. Consider using an end-of-epoch checkpoint + diff --git a/outputs/2026-06-25/21-29-00/.hydra/config.yaml b/outputs/2026-06-25/21-29-00/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d24738b99d91e3ae14cbbe8a4c211bcf4b2fc76b --- /dev/null +++ b/outputs/2026-06-25/21-29-00/.hydra/config.yaml @@ -0,0 +1,211 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 60000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 700 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 5000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + noise_route: + anchor: all + dynamic: all + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 20000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-25/21-29-00/.hydra/hydra.yaml b/outputs/2026-06-25/21-29-00/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4067908c5fe8b79bcd68caab28d333abcbaebb08 --- /dev/null +++ b/outputs/2026-06-25/21-29-00/.hydra/hydra.yaml @@ -0,0 +1,216 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=700 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=5000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=20000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=all + - algorithm.noise_route.dynamic=all + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_max_fraction=0.50 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_max_fraction=0.50,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.n_tokens=8,algorithm.noise_route.anchor=all,algorithm.noise_route.dynamic=all,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=20000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=5000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=700,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-25/21-29-00 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-25/21-29-00/.hydra/overrides.yaml b/outputs/2026-06-25/21-29-00/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4b85025abd8fab687599f0ac1ed0ad4c8ddb9f9a --- /dev/null +++ b/outputs/2026-06-25/21-29-00/.hydra/overrides.yaml @@ -0,0 +1,58 @@ +- +name=train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=700 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=5000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=20000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=all +- algorithm.noise_route.dynamic=all +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_max_fraction=0.50 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-06-25/21-29-00/main.log b/outputs/2026-06-25/21-29-00/main.log new file mode 100644 index 0000000000000000000000000000000000000000..f7520d551c52b52c7a0d750a73fdb606ba34ea3c --- /dev/null +++ b/outputs/2026-06-25/21-29-00/main.log @@ -0,0 +1,9 @@ +[2026-06-25 21:30:32,157][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-25 21:30:56,337][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-25 21:30:58,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 14:37:00,534][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-06-26 14:38:39,396][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-06-26 14:39:01,295][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 14:39:03,399][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-26/21-54-56/.hydra/config.yaml b/outputs/2026-06-26/21-54-56/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7c97550b32338ca18b91468ee081745e758dd303 --- /dev/null +++ b/outputs/2026-06-26/21-54-56/.hydra/config.yaml @@ -0,0 +1,211 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 60000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 700 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: recent + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 5000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 20000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-26/21-54-56/.hydra/hydra.yaml b/outputs/2026-06-26/21-54-56/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a6b58f77abfd2e100b4c099608e303f2f5e58d4e --- /dev/null +++ b/outputs/2026-06-26/21-54-56/.hydra/hydra.yaml @@ -0,0 +1,216 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=700 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=recent + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=5000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=20000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_max_fraction=0.50 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_max_fraction=0.50,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=20000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=5000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.dynamic.selection_policy=recent,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=700,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-26/21-54-56 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-26/21-54-56/.hydra/overrides.yaml b/outputs/2026-06-26/21-54-56/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..634c9c262ff8c9a6fac58a418e60d4661df1695a --- /dev/null +++ b/outputs/2026-06-26/21-54-56/.hydra/overrides.yaml @@ -0,0 +1,58 @@ +- +name=train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=700 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=recent +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=5000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=20000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_max_fraction=0.50 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-06-26/21-54-56/main.log b/outputs/2026-06-26/21-54-56/main.log new file mode 100644 index 0000000000000000000000000000000000000000..4ba56b169df0b8bc2bf8919af62a33d1b08c928c --- /dev/null +++ b/outputs/2026-06-26/21-54-56/main.log @@ -0,0 +1,25 @@ +[2026-06-26 21:55:58,068][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-26 21:55:58,070][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-26 21:55:58,084][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-26 21:55:58,098][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-26 21:55:58,106][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-26 21:56:44,104][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:56:44,112][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:56:44,124][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:56:44,136][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:56:44,226][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:56:49,709][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:56:49,709][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:56:49,709][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:56:49,709][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:56:49,709][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:58:52,480][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:58:52,482][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:58:52,484][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:58:52,489][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:58:52,533][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:58:54,484][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:58:54,484][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:58:54,484][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:58:54,484][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:58:54,484][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-26/21-54-57/.hydra/config.yaml b/outputs/2026-06-26/21-54-57/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7c97550b32338ca18b91468ee081745e758dd303 --- /dev/null +++ b/outputs/2026-06-26/21-54-57/.hydra/config.yaml @@ -0,0 +1,211 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 60000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 700 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: recent + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 5000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 20000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-26/21-54-57/.hydra/hydra.yaml b/outputs/2026-06-26/21-54-57/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..43d9cc116978e091425c5889505fed36a915306f --- /dev/null +++ b/outputs/2026-06-26/21-54-57/.hydra/hydra.yaml @@ -0,0 +1,216 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=700 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=recent + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=5000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=20000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_max_fraction=0.50 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_max_fraction=0.50,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=20000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=5000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.dynamic.selection_policy=recent,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=700,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-26/21-54-57 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-26/21-54-57/.hydra/overrides.yaml b/outputs/2026-06-26/21-54-57/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..634c9c262ff8c9a6fac58a418e60d4661df1695a --- /dev/null +++ b/outputs/2026-06-26/21-54-57/.hydra/overrides.yaml @@ -0,0 +1,58 @@ +- +name=train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=700 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=recent +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=5000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=20000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_max_fraction=0.50 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-06-26/21-54-57/main.log b/outputs/2026-06-26/21-54-57/main.log new file mode 100644 index 0000000000000000000000000000000000000000..ac44034ee605512e0efcf6519b1f71ae51b043f2 --- /dev/null +++ b/outputs/2026-06-26/21-54-57/main.log @@ -0,0 +1,68 @@ +[2026-06-26 21:55:29,654][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-26 21:55:29,778][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-26 21:55:29,778][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-26 21:55:29,778][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-26 21:55:29,778][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-26 21:55:29,779][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-26 21:55:58,064][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-26 21:55:58,079][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-26 21:55:58,127][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-26 21:56:05,222][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-26 21:56:20,976][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 21:56:44,107][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:56:44,127][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:56:44,131][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:56:47,599][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-26 21:56:48,651][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 21:56:49,709][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:56:49,709][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:56:49,709][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:56:51,227][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=60000` reached. +[2026-06-26 21:57:14,412][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-26 21:57:15,913][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-26 21:57:15,914][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-26 21:57:15,914][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-26 21:57:15,914][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-26 21:57:15,914][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-26 21:58:29,958][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 21:58:52,478][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:58:52,479][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:58:52,488][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:58:53,295][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-26 21:58:54,456][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_base_recent_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 21:58:54,484][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:58:54,484][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:58:54,484][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 22:01:33,780][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-26 22:01:33,793][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-26 22:01:33,794][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-26 22:01:33,794][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-28 15:39:11,810][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=200000` reached. diff --git a/outputs/2026-06-26/21-55-27/.hydra/config.yaml b/outputs/2026-06-26/21-55-27/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7021a9adaba2f265eb735cd2e1ba34e1b0e98274 --- /dev/null +++ b/outputs/2026-06-26/21-55-27/.hydra/config.yaml @@ -0,0 +1,211 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 60000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 700 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 5000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 20000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-26/21-55-27/.hydra/hydra.yaml b/outputs/2026-06-26/21-55-27/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4a7e587cecad34f3002453281110566ea0b7b149 --- /dev/null +++ b/outputs/2026-06-26/21-55-27/.hydra/hydra.yaml @@ -0,0 +1,216 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=700 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=5000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=20000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_max_fraction=0.50 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_max_fraction=0.50,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=20000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=5000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=700,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-26/21-55-27 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-26/21-55-27/.hydra/overrides.yaml b/outputs/2026-06-26/21-55-27/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..50839acd41803ee9a5606af6358b422484826767 --- /dev/null +++ b/outputs/2026-06-26/21-55-27/.hydra/overrides.yaml @@ -0,0 +1,58 @@ +- +name=train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=700 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=5000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=20000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_max_fraction=0.50 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-06-26/21-55-27/main.log b/outputs/2026-06-26/21-55-27/main.log new file mode 100644 index 0000000000000000000000000000000000000000..21c0c418b63718bca8debeb4909d6a9a5a5c909b --- /dev/null +++ b/outputs/2026-06-26/21-55-27/main.log @@ -0,0 +1,93 @@ +[2026-06-26 21:56:15,375][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-26 21:56:15,513][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-26 21:56:15,514][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-26 21:56:15,514][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-26 21:56:15,514][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-26 21:56:15,514][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-26 21:57:03,263][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-26 21:57:03,264][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-26 21:57:03,268][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-26 21:57:03,269][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-26 21:57:03,273][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-26 21:57:03,286][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-26 21:57:03,287][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-26 21:57:03,295][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-26 21:57:08,860][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-26 21:57:28,216][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 21:57:48,647][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:57:48,651][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:57:48,657][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:57:48,663][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:57:48,668][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:57:48,670][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:57:48,754][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:57:48,808][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 21:57:50,706][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-26 21:57:51,793][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 21:57:52,794][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:57:52,794][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:57:52,794][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:57:52,794][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:57:52,794][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:57:52,794][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:57:52,794][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:57:52,794][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 21:57:54,645][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=60000` reached. +[2026-06-26 21:58:17,968][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-26 21:58:20,252][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-26 21:58:20,253][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-26 21:58:20,253][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-26 21:58:20,253][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-26 21:58:20,253][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-26 22:00:06,501][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 22:00:28,913][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 22:00:28,917][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 22:00:28,924][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 22:00:28,929][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 22:00:28,932][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 22:00:28,933][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 22:00:28,936][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 22:00:28,937][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 22:00:29,725][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-26 22:00:30,865][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 22:00:30,893][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 22:00:30,893][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 22:00:30,893][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 22:00:30,893][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 22:00:30,893][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 22:00:30,893][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 22:00:30,893][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 22:00:30,893][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 22:03:14,646][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-26 22:03:14,660][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-26 22:03:14,660][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-26 22:03:14,660][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-28 15:38:08,155][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=200000` reached. diff --git a/outputs/2026-06-26/23-03-39/.hydra/config.yaml b/outputs/2026-06-26/23-03-39/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3e0d1eacb4a756bcfda7d47203d6772fdda4254a --- /dev/null +++ b/outputs/2026-06-26/23-03-39/.hydra/config.yaml @@ -0,0 +1,211 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 60000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 700 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 5000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: false + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 20000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-26/23-03-39/.hydra/hydra.yaml b/outputs/2026-06-26/23-03-39/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..99bd6011eb8e1d54ff499c9af5654c318fe705d0 --- /dev/null +++ b/outputs/2026-06-26/23-03-39/.hydra/hydra.yaml @@ -0,0 +1,216 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=700 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=5000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=20000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=false + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_max_fraction=0.50 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_max_fraction=0.50,algorithm.memory_noise.enabled=false,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=20000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=5000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=700,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-26/23-03-39 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-26/23-03-39/.hydra/overrides.yaml b/outputs/2026-06-26/23-03-39/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5c8056a58520299ba8204bad4bbcc640e272ce00 --- /dev/null +++ b/outputs/2026-06-26/23-03-39/.hydra/overrides.yaml @@ -0,0 +1,58 @@ +- +name=train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=700 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=5000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=20000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=false +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_max_fraction=0.50 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-06-26/23-03-39/main.log b/outputs/2026-06-26/23-03-39/main.log new file mode 100644 index 0000000000000000000000000000000000000000..cfb455ec88a48f8f76ff1d5cd0cc83fc2ba3a6b8 --- /dev/null +++ b/outputs/2026-06-26/23-03-39/main.log @@ -0,0 +1,5 @@ +[2026-06-26 23:05:24,503][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-26 23:06:11,327][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 23:06:15,383][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 23:08:33,428][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 23:08:36,001][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-26/23-03-40/.hydra/config.yaml b/outputs/2026-06-26/23-03-40/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3e0d1eacb4a756bcfda7d47203d6772fdda4254a --- /dev/null +++ b/outputs/2026-06-26/23-03-40/.hydra/config.yaml @@ -0,0 +1,211 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 60000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 700 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 5000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: false + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 20000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-26/23-03-40/.hydra/hydra.yaml b/outputs/2026-06-26/23-03-40/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..78be1f76e6f591b414b5ad4646fd03f4f830b31e --- /dev/null +++ b/outputs/2026-06-26/23-03-40/.hydra/hydra.yaml @@ -0,0 +1,216 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=700 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=5000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=20000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=false + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_max_fraction=0.50 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_max_fraction=0.50,algorithm.memory_noise.enabled=false,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=20000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=5000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=700,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-26/23-03-40 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-26/23-03-40/.hydra/overrides.yaml b/outputs/2026-06-26/23-03-40/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5c8056a58520299ba8204bad4bbcc640e272ce00 --- /dev/null +++ b/outputs/2026-06-26/23-03-40/.hydra/overrides.yaml @@ -0,0 +1,58 @@ +- +name=train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=700 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=5000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=20000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=false +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_max_fraction=0.50 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-06-26/23-03-40/main.log b/outputs/2026-06-26/23-03-40/main.log new file mode 100644 index 0000000000000000000000000000000000000000..ac3c1955fe34eb353ad8b2b6b11339ea7c2de678 --- /dev/null +++ b/outputs/2026-06-26/23-03-40/main.log @@ -0,0 +1,88 @@ +[2026-06-26 23:04:33,447][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-26 23:04:33,664][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-26 23:04:33,664][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-26 23:04:33,664][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-26 23:04:33,664][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-26 23:04:33,665][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-26 23:05:24,505][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-26 23:05:24,507][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-26 23:05:24,518][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-26 23:05:24,527][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-26 23:05:24,529][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-26 23:05:24,539][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-26 23:05:24,543][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-26 23:05:30,498][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-26 23:05:49,112][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 23:06:11,289][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 23:06:11,308][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 23:06:11,320][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 23:06:11,341][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 23:06:11,346][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 23:06:11,348][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 23:06:11,382][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 23:06:13,352][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-26 23:06:14,390][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 23:06:15,383][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 23:06:15,383][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 23:06:15,383][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 23:06:15,383][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 23:06:15,383][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 23:06:15,383][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 23:06:15,383][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 23:06:16,988][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=60000` reached. +[2026-06-26 23:06:40,779][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-26 23:06:41,977][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-26 23:06:41,978][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-26 23:06:41,978][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-26 23:06:41,978][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-26 23:06:41,978][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-26 23:08:27,478][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 23:08:33,430][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 23:08:33,611][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 23:08:33,612][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 23:08:33,613][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 23:08:33,616][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 23:08:33,694][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 23:08:34,015][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-26 23:08:34,736][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-26 23:08:35,995][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_memory_noise_a2d4r2_route_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-26 23:08:36,001][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 23:08:36,001][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 23:08:36,001][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 23:08:36,001][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 23:08:36,001][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 23:08:36,001][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 23:08:36,001][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-26 23:11:13,621][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-26 23:11:13,633][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-26 23:11:13,633][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-26 23:11:13,633][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-28 16:32:53,866][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=200000` reached. diff --git a/outputs/2026-06-27/02-41-42/.hydra/config.yaml b/outputs/2026-06-27/02-41-42/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d24738b99d91e3ae14cbbe8a4c211bcf4b2fc76b --- /dev/null +++ b/outputs/2026-06-27/02-41-42/.hydra/config.yaml @@ -0,0 +1,211 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 60000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 700 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 5000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + noise_route: + anchor: all + dynamic: all + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 20000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-27/02-41-42/.hydra/hydra.yaml b/outputs/2026-06-27/02-41-42/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..38441838abbd89c2506a29ad3c6bd51cc3f0f514 --- /dev/null +++ b/outputs/2026-06-27/02-41-42/.hydra/hydra.yaml @@ -0,0 +1,216 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=700 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=5000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=20000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=all + - algorithm.noise_route.dynamic=all + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_max_fraction=0.50 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_max_fraction=0.50,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.n_tokens=8,algorithm.noise_route.anchor=all,algorithm.noise_route.dynamic=all,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=20000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=5000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=700,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-27/02-41-42 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-27/02-41-42/.hydra/overrides.yaml b/outputs/2026-06-27/02-41-42/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4b85025abd8fab687599f0ac1ed0ad4c8ddb9f9a --- /dev/null +++ b/outputs/2026-06-27/02-41-42/.hydra/overrides.yaml @@ -0,0 +1,58 @@ +- +name=train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=700 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=5000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=20000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=all +- algorithm.noise_route.dynamic=all +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_max_fraction=0.50 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-06-27/02-41-42/main.log b/outputs/2026-06-27/02-41-42/main.log new file mode 100644 index 0000000000000000000000000000000000000000..cb3acb8703a578c3df3e2cea30030bcbdd7b4fa5 --- /dev/null +++ b/outputs/2026-06-27/02-41-42/main.log @@ -0,0 +1,5 @@ +[2026-06-27 02:43:17,623][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-27 02:44:04,956][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-27 02:44:09,717][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-27 02:46:44,479][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-27 02:46:46,423][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-27/02-41-43/.hydra/config.yaml b/outputs/2026-06-27/02-41-43/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d24738b99d91e3ae14cbbe8a4c211bcf4b2fc76b --- /dev/null +++ b/outputs/2026-06-27/02-41-43/.hydra/config.yaml @@ -0,0 +1,211 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 5000 + every_n_epochs: null + train_time_interval: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 60000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 5000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 700 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 5000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + noise_route: + anchor: all + dynamic: all + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 20000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-27/02-41-43/.hydra/hydra.yaml b/outputs/2026-06-27/02-41-43/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..147c657e131c904cf58b8f5549101a7a9630614e --- /dev/null +++ b/outputs/2026-06-27/02-41-43/.hydra/hydra.yaml @@ -0,0 +1,216 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=700 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=5000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=20000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=all + - algorithm.noise_route.dynamic=all + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_max_fraction=0.50 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=5000 + - experiment.validation.val_every_n_step=5000 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_max_fraction=0.50,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.n_tokens=8,algorithm.noise_route.anchor=all,algorithm.noise_route.dynamic=all,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=20000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=5000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=700,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=5000,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=5000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-27/02-41-43 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-27/02-41-43/.hydra/overrides.yaml b/outputs/2026-06-27/02-41-43/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4b85025abd8fab687599f0ac1ed0ad4c8ddb9f9a --- /dev/null +++ b/outputs/2026-06-27/02-41-43/.hydra/overrides.yaml @@ -0,0 +1,58 @@ +- +name=train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=700 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=5000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=20000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=all +- algorithm.noise_route.dynamic=all +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_max_fraction=0.50 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:60000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=5000 +- experiment.validation.val_every_n_step=5000 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-06-27/02-41-43/main.log b/outputs/2026-06-27/02-41-43/main.log new file mode 100644 index 0000000000000000000000000000000000000000..9018c8fd5d634eee9192b2d811f1db72ab3bfc45 --- /dev/null +++ b/outputs/2026-06-27/02-41-43/main.log @@ -0,0 +1,87 @@ +[2026-06-27 02:42:28,496][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-27 02:42:28,648][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-27 02:42:28,649][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-27 02:42:28,649][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-27 02:42:28,649][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-27 02:42:28,649][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-27 02:43:17,576][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-27 02:43:17,577][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-27 02:43:17,593][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-27 02:43:17,601][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-27 02:43:17,612][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-27 02:43:17,614][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-27 02:43:17,621][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-27 02:43:23,456][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-27 02:43:42,740][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-27 02:44:04,944][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-27 02:44:04,949][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-27 02:44:04,951][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-27 02:44:04,951][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-27 02:44:04,952][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-27 02:44:04,954][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-27 02:44:04,958][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-27 02:44:06,715][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-27 02:44:07,776][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-27 02:44:09,717][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-27 02:44:09,717][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-27 02:44:09,717][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-27 02:44:09,717][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-27 02:44:09,717][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-27 02:44:09,717][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-27 02:44:09,717][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-27 02:44:11,952][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=60000` reached. +[2026-06-27 02:44:34,170][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-27 02:44:34,986][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-27 02:44:34,986][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-27 02:44:34,986][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-27 02:44:34,986][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-27 02:44:34,986][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-27 02:46:21,004][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-27 02:46:44,465][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-27 02:46:44,466][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-27 02:46:44,469][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-27 02:46:44,470][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-27 02:46:44,471][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-27 02:46:44,472][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-27 02:46:44,474][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-27 02:46:45,267][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-27 02:46:46,411][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_a2d4r2_ntok8_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_step60000.ckpt +[2026-06-27 02:46:46,422][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-27 02:46:46,422][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-27 02:46:46,422][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-27 02:46:46,422][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-27 02:46:46,422][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-27 02:46:46,422][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-27 02:46:46,422][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-27 02:49:25,621][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-27 02:49:26,026][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-27 02:49:26,027][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-27 02:49:26,027][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + diff --git a/outputs/2026-06-30/19-06-53/.hydra/config.yaml b/outputs/2026-06-30/19-06-53/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c81ef2c4af5ee421ba6f3a9526a388b050457bab --- /dev/null +++ b/outputs/2026-06-30/19-06-53/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 150000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 10 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 30000 + dataset: + wo_updown: true + - name: with_updown + until_step: 150000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.0 + dynamic_max_fraction: 0.25 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-30/19-06-53/.hydra/hydra.yaml b/outputs/2026-06-30/19-06-53/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4e075c4e1c6082fa626b19bb8159fccc31817528 --- /dev/null +++ b/outputs/2026-06-30/19-06-53/.hydra/hydra.yaml @@ -0,0 +1,219 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=1000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.0 + - algorithm.memory_noise.dynamic_max_fraction=0.25 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=10 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=150000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.0,algorithm.memory_noise.dynamic_max_fraction=0.25,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=1000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=10,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]},experiment.training.max_steps=150000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-30/19-06-53 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-30/19-06-53/.hydra/overrides.yaml b/outputs/2026-06-30/19-06-53/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..848c831fb4d100b4cb3423a74c6f383fd35a56fb --- /dev/null +++ b/outputs/2026-06-30/19-06-53/.hydra/overrides.yaml @@ -0,0 +1,61 @@ +- +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=1000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.0 +- algorithm.memory_noise.dynamic_max_fraction=0.25 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=10 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=150000 diff --git a/outputs/2026-06-30/19-06-53/main.log b/outputs/2026-06-30/19-06-53/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/2026-06-30/19-06-54/.hydra/config.yaml b/outputs/2026-06-30/19-06-54/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c81ef2c4af5ee421ba6f3a9526a388b050457bab --- /dev/null +++ b/outputs/2026-06-30/19-06-54/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 150000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 10 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 30000 + dataset: + wo_updown: true + - name: with_updown + until_step: 150000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.0 + dynamic_max_fraction: 0.25 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-30/19-06-54/.hydra/hydra.yaml b/outputs/2026-06-30/19-06-54/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fbff5f53736958c46391aa3e9fbd37e586f14d8c --- /dev/null +++ b/outputs/2026-06-30/19-06-54/.hydra/hydra.yaml @@ -0,0 +1,219 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=1000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.0 + - algorithm.memory_noise.dynamic_max_fraction=0.25 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=10 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=150000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.0,algorithm.memory_noise.dynamic_max_fraction=0.25,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=1000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=10,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]},experiment.training.max_steps=150000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-30/19-06-54 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-30/19-06-54/.hydra/overrides.yaml b/outputs/2026-06-30/19-06-54/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..848c831fb4d100b4cb3423a74c6f383fd35a56fb --- /dev/null +++ b/outputs/2026-06-30/19-06-54/.hydra/overrides.yaml @@ -0,0 +1,61 @@ +- +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=1000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.0 +- algorithm.memory_noise.dynamic_max_fraction=0.25 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=10 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=150000 diff --git a/outputs/2026-06-30/19-06-54/main.log b/outputs/2026-06-30/19-06-54/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/2026-06-30/19-06-55/.hydra/config.yaml b/outputs/2026-06-30/19-06-55/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c81ef2c4af5ee421ba6f3a9526a388b050457bab --- /dev/null +++ b/outputs/2026-06-30/19-06-55/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 150000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 10 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 30000 + dataset: + wo_updown: true + - name: with_updown + until_step: 150000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.0 + dynamic_max_fraction: 0.25 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-30/19-06-55/.hydra/hydra.yaml b/outputs/2026-06-30/19-06-55/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a2e872d614bcbfb9dac5e75ba0060d58c51edc64 --- /dev/null +++ b/outputs/2026-06-30/19-06-55/.hydra/hydra.yaml @@ -0,0 +1,219 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=1000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.0 + - algorithm.memory_noise.dynamic_max_fraction=0.25 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=10 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=150000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.0,algorithm.memory_noise.dynamic_max_fraction=0.25,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=1000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=10,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]},experiment.training.max_steps=150000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-30/19-06-55 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-30/19-06-55/.hydra/overrides.yaml b/outputs/2026-06-30/19-06-55/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..848c831fb4d100b4cb3423a74c6f383fd35a56fb --- /dev/null +++ b/outputs/2026-06-30/19-06-55/.hydra/overrides.yaml @@ -0,0 +1,61 @@ +- +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=1000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.0 +- algorithm.memory_noise.dynamic_max_fraction=0.25 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=10 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=150000 diff --git a/outputs/2026-06-30/19-06-55/main.log b/outputs/2026-06-30/19-06-55/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/2026-06-30/19-06-59/.hydra/config.yaml b/outputs/2026-06-30/19-06-59/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..44622f84726ec574fca73aa589ee22826a9d62a2 --- /dev/null +++ b/outputs/2026-06-30/19-06-59/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 150000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 10 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 30000 + dataset: + wo_updown: true + - name: with_updown + until_step: 150000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: false + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.0 + dynamic_max_fraction: 0.25 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-30/19-06-59/.hydra/hydra.yaml b/outputs/2026-06-30/19-06-59/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..52889d5ce0492fce5302e227f4e2834f47b46a45 --- /dev/null +++ b/outputs/2026-06-30/19-06-59/.hydra/hydra.yaml @@ -0,0 +1,219 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=false + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=1000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.0 + - algorithm.memory_noise.dynamic_max_fraction=0.25 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=10 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=150000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.0,algorithm.memory_noise.dynamic_max_fraction=0.25,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=1000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.causal=false,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=10,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]},experiment.training.max_steps=150000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-30/19-06-59 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-30/19-06-59/.hydra/overrides.yaml b/outputs/2026-06-30/19-06-59/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..efdd985de2643ea464ba05109c047aedfa2b3a80 --- /dev/null +++ b/outputs/2026-06-30/19-06-59/.hydra/overrides.yaml @@ -0,0 +1,61 @@ +- +name=train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=false +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=1000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.0 +- algorithm.memory_noise.dynamic_max_fraction=0.25 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=10 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=150000 diff --git a/outputs/2026-06-30/19-06-59/main.log b/outputs/2026-06-30/19-06-59/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/2026-06-30/19-07-00/.hydra/config.yaml b/outputs/2026-06-30/19-07-00/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..44622f84726ec574fca73aa589ee22826a9d62a2 --- /dev/null +++ b/outputs/2026-06-30/19-07-00/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 150000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 10 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 30000 + dataset: + wo_updown: true + - name: with_updown + until_step: 150000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: false + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.0 + dynamic_max_fraction: 0.25 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-30/19-07-00/.hydra/hydra.yaml b/outputs/2026-06-30/19-07-00/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d8eb191b6cad90e142aa8898e21446de2e37b1cf --- /dev/null +++ b/outputs/2026-06-30/19-07-00/.hydra/hydra.yaml @@ -0,0 +1,219 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=false + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=1000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.0 + - algorithm.memory_noise.dynamic_max_fraction=0.25 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=10 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=150000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.0,algorithm.memory_noise.dynamic_max_fraction=0.25,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=1000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.causal=false,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=10,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]},experiment.training.max_steps=150000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-30/19-07-00 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-30/19-07-00/.hydra/overrides.yaml b/outputs/2026-06-30/19-07-00/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..efdd985de2643ea464ba05109c047aedfa2b3a80 --- /dev/null +++ b/outputs/2026-06-30/19-07-00/.hydra/overrides.yaml @@ -0,0 +1,61 @@ +- +name=train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=false +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=1000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.0 +- algorithm.memory_noise.dynamic_max_fraction=0.25 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=10 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=150000 diff --git a/outputs/2026-06-30/19-07-00/main.log b/outputs/2026-06-30/19-07-00/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/2026-06-30/19-07-07/.hydra/config.yaml b/outputs/2026-06-30/19-07-07/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8ff5d963c0848b9bb2ea4f52dadab7f3ccf650b6 --- /dev/null +++ b/outputs/2026-06-30/19-07-07/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 150000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 10 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 30000 + dataset: + wo_updown: true + - name: with_updown + until_step: 150000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.0 + dynamic_max_fraction: 0.25 + revisit_max_fraction: 0.25 + noise_route: + anchor: all + dynamic: all + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-30/19-07-07/.hydra/hydra.yaml b/outputs/2026-06-30/19-07-07/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..71a9e21e21314903c497f5ca1693847b81cf9c57 --- /dev/null +++ b/outputs/2026-06-30/19-07-07/.hydra/hydra.yaml @@ -0,0 +1,219 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=1000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=all + - algorithm.noise_route.dynamic=all + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.0 + - algorithm.memory_noise.dynamic_max_fraction=0.25 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=10 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=150000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.0,algorithm.memory_noise.dynamic_max_fraction=0.25,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=all,algorithm.noise_route.dynamic=all,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=1000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=10,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]},experiment.training.max_steps=150000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-30/19-07-07 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-30/19-07-07/.hydra/overrides.yaml b/outputs/2026-06-30/19-07-07/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7e3f2353d03287af5ccbd829323aa8d3239cc6e1 --- /dev/null +++ b/outputs/2026-06-30/19-07-07/.hydra/overrides.yaml @@ -0,0 +1,61 @@ +- +name=train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=1000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=all +- algorithm.noise_route.dynamic=all +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.0 +- algorithm.memory_noise.dynamic_max_fraction=0.25 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=10 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=150000 diff --git a/outputs/2026-06-30/19-07-07/main.log b/outputs/2026-06-30/19-07-07/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/2026-06-30/19-23-38/.hydra/config.yaml b/outputs/2026-06-30/19-23-38/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c81ef2c4af5ee421ba6f3a9526a388b050457bab --- /dev/null +++ b/outputs/2026-06-30/19-23-38/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 150000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 10 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 30000 + dataset: + wo_updown: true + - name: with_updown + until_step: 150000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.0 + dynamic_max_fraction: 0.25 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-30/19-23-38/.hydra/hydra.yaml b/outputs/2026-06-30/19-23-38/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7cb9ab94762f643ca12f7da3bb6ccc96cfd46c2c --- /dev/null +++ b/outputs/2026-06-30/19-23-38/.hydra/hydra.yaml @@ -0,0 +1,219 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=1000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.0 + - algorithm.memory_noise.dynamic_max_fraction=0.25 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=10 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=150000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.0,algorithm.memory_noise.dynamic_max_fraction=0.25,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=1000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=10,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]},experiment.training.max_steps=150000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-30/19-23-38 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-30/19-23-38/.hydra/overrides.yaml b/outputs/2026-06-30/19-23-38/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..848c831fb4d100b4cb3423a74c6f383fd35a56fb --- /dev/null +++ b/outputs/2026-06-30/19-23-38/.hydra/overrides.yaml @@ -0,0 +1,61 @@ +- +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=1000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.0 +- algorithm.memory_noise.dynamic_max_fraction=0.25 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=10 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=150000 diff --git a/outputs/2026-06-30/19-23-38/main.log b/outputs/2026-06-30/19-23-38/main.log new file mode 100644 index 0000000000000000000000000000000000000000..4be82153d1cd1e929fe2bb9083142f9f2b4408b9 --- /dev/null +++ b/outputs/2026-06-30/19-23-38/main.log @@ -0,0 +1,155 @@ +[2026-06-30 19:24:28,849][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-06-30 19:24:28,850][pytorch_lightning.utilities.rank_zero][INFO] - Missing checkpoint keys: timestamp_embedding.mlp.0.weight, timestamp_embedding.mlp.0.bias, timestamp_embedding.mlp.2.weight, timestamp_embedding.mlp.2.bias, spatial_rotary_emb.freqs, temporal_rotary_emb.freqs, pose_embedder.model.0.weight, pose_embedder.model.0.bias, ... (+560 more) +[2026-06-30 19:24:28,939][pytorch_lightning.utilities.rank_zero][INFO] - Zeroed frame-memory gate slices for base checkpoint load: blocks.0.r_adaLN_modulation.1.weight, blocks.0.r_adaLN_modulation.1.bias, blocks.1.r_adaLN_modulation.1.weight, blocks.1.r_adaLN_modulation.1.bias, blocks.2.r_adaLN_modulation.1.weight, blocks.2.r_adaLN_modulation.1.bias, blocks.3.r_adaLN_modulation.1.weight, blocks.3.r_adaLN_modulation.1.bias, ... (+24 more) +[2026-06-30 19:24:28,939][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: 302 tensors. +[2026-06-30 19:24:29,025][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-06-30 19:24:29,062][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: 228 tensors. +[2026-06-30 19:24:29,107][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-30 19:24:29,256][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-30 19:24:29,257][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-30 19:24:29,257][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-30 19:24:29,257][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-30 19:24:29,257][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-30 19:25:02,079][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-30 19:25:02,079][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-30 19:25:02,079][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-30 19:25:02,080][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-30 19:25:02,183][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-30 19:25:02,780][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-30 19:25:08,639][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-30 19:25:14,146][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:14,146][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:14,146][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:14,146][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:14,146][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:14,146][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:17,764][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-30 19:25:17,769][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 19:25:17,769][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 19:25:17,769][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 19:25:17,769][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 19:25:17,769][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 19:25:17,769][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 20:07:43,008][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,149][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,149][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,149][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,149][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,149][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,150][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,150][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,150][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,150][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,150][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,150][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,150][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,150][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,150][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,151][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,151][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,151][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,151][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,151][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,151][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,151][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,151][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:07:56,152][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-01 06:32:04,534][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=30000` reached. +[2026-07-01 06:32:26,101][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-01 06:32:26,102][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:32:26,102][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:32:26,102][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:32:26,102][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:32:26,102][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:32:26,102][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:32:26,618][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-01 06:32:26,618][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-01 06:32:26,618][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-01 06:32:26,618][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-01 06:32:26,619][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-01 06:34:09,649][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/callbacks/model_checkpoint.py:639: Checkpoint directory /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/checkpoints exists and is not empty. + +[2026-07-01 06:34:09,649][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/checkpoints/curriculum_01_no_updown_step30000.ckpt +[2026-07-01 06:34:09,660][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:34:09,662][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:34:09,664][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:34:09,667][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:34:09,670][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:34:09,673][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:34:15,865][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:34:16,198][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:34:16,198][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:34:16,273][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:34:16,464][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:34:16,569][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:34:17,371][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-07-01 06:34:18,560][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/checkpoints/curriculum_01_no_updown_step30000.ckpt +[2026-07-01 06:34:18,606][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:34:18,606][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:34:18,606][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:34:18,606][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:34:18,606][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:34:18,606][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:34:19,846][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/loops/training_epoch_loop.py:154: You're resuming from a checkpoint that ended before the epoch ended. This can cause unreliable results if further training is done. Consider using an end-of-epoch checkpoint + +[2026-07-03 05:19:16,315][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=150000` reached. diff --git a/outputs/2026-06-30/19-23-39/.hydra/config.yaml b/outputs/2026-06-30/19-23-39/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c81ef2c4af5ee421ba6f3a9526a388b050457bab --- /dev/null +++ b/outputs/2026-06-30/19-23-39/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 150000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 10 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 30000 + dataset: + wo_updown: true + - name: with_updown + until_step: 150000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.0 + dynamic_max_fraction: 0.25 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-30/19-23-39/.hydra/hydra.yaml b/outputs/2026-06-30/19-23-39/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cafbf07f8f5322dd6f394dc43d87257559ab074e --- /dev/null +++ b/outputs/2026-06-30/19-23-39/.hydra/hydra.yaml @@ -0,0 +1,219 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=1000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.0 + - algorithm.memory_noise.dynamic_max_fraction=0.25 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=10 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=150000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.0,algorithm.memory_noise.dynamic_max_fraction=0.25,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=1000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=10,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]},experiment.training.max_steps=150000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-30/19-23-39 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-30/19-23-39/.hydra/overrides.yaml b/outputs/2026-06-30/19-23-39/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..848c831fb4d100b4cb3423a74c6f383fd35a56fb --- /dev/null +++ b/outputs/2026-06-30/19-23-39/.hydra/overrides.yaml @@ -0,0 +1,61 @@ +- +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=1000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.0 +- algorithm.memory_noise.dynamic_max_fraction=0.25 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=10 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=150000 diff --git a/outputs/2026-06-30/19-23-39/main.log b/outputs/2026-06-30/19-23-39/main.log new file mode 100644 index 0000000000000000000000000000000000000000..afbae609ba738f868915291cef99a4cd16618636 --- /dev/null +++ b/outputs/2026-06-30/19-23-39/main.log @@ -0,0 +1,18 @@ +[2026-06-30 19:25:02,080][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-30 19:25:02,080][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-30 19:25:14,146][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:14,146][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:17,769][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 19:25:17,769][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:32:26,102][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:32:26,102][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:34:09,668][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:34:09,670][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:34:15,975][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:34:16,467][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:34:18,606][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:34:18,606][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-30/19-23-44/.hydra/config.yaml b/outputs/2026-06-30/19-23-44/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..44622f84726ec574fca73aa589ee22826a9d62a2 --- /dev/null +++ b/outputs/2026-06-30/19-23-44/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 150000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 10 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 30000 + dataset: + wo_updown: true + - name: with_updown + until_step: 150000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: false + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.0 + dynamic_max_fraction: 0.25 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-30/19-23-44/.hydra/hydra.yaml b/outputs/2026-06-30/19-23-44/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ecb036598152760cb4e0ee17aafbe9c0ac36f726 --- /dev/null +++ b/outputs/2026-06-30/19-23-44/.hydra/hydra.yaml @@ -0,0 +1,219 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=false + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=1000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.0 + - algorithm.memory_noise.dynamic_max_fraction=0.25 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=10 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=150000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.0,algorithm.memory_noise.dynamic_max_fraction=0.25,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=1000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.causal=false,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=10,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]},experiment.training.max_steps=150000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-30/19-23-44 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-30/19-23-44/.hydra/overrides.yaml b/outputs/2026-06-30/19-23-44/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..efdd985de2643ea464ba05109c047aedfa2b3a80 --- /dev/null +++ b/outputs/2026-06-30/19-23-44/.hydra/overrides.yaml @@ -0,0 +1,61 @@ +- +name=train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=false +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=1000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.0 +- algorithm.memory_noise.dynamic_max_fraction=0.25 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=10 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=150000 diff --git a/outputs/2026-06-30/19-23-44/main.log b/outputs/2026-06-30/19-23-44/main.log new file mode 100644 index 0000000000000000000000000000000000000000..c0cb114a5f76aab0dd2bd8a76d01a867040f15ae --- /dev/null +++ b/outputs/2026-06-30/19-23-44/main.log @@ -0,0 +1,29 @@ +[2026-06-30 19:25:10,901][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-30 19:25:10,903][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-30 19:25:10,904][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-30 19:25:23,015][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:23,015][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:23,015][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:25,769][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 19:25:25,769][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 19:25:25,769][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:33:24,811][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:33:24,812][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:33:24,812][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:35:15,565][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:35:15,578][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:35:15,580][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:35:34,844][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:35:34,846][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:35:34,858][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:35:36,897][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:35:36,897][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:35:36,897][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 23:51:08,644][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-07-01 23:51:08,645][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-30/19-23-45/.hydra/config.yaml b/outputs/2026-06-30/19-23-45/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..44622f84726ec574fca73aa589ee22826a9d62a2 --- /dev/null +++ b/outputs/2026-06-30/19-23-45/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 150000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 10 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 30000 + dataset: + wo_updown: true + - name: with_updown + until_step: 150000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: false + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.0 + dynamic_max_fraction: 0.25 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-30/19-23-45/.hydra/hydra.yaml b/outputs/2026-06-30/19-23-45/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..af7cbc1c62907cf1ce5f6625049f0d06d8b21f4c --- /dev/null +++ b/outputs/2026-06-30/19-23-45/.hydra/hydra.yaml @@ -0,0 +1,219 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=false + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=1000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.0 + - algorithm.memory_noise.dynamic_max_fraction=0.25 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=10 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=150000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.0,algorithm.memory_noise.dynamic_max_fraction=0.25,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=1000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.causal=false,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=10,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]},experiment.training.max_steps=150000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-30/19-23-45 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-30/19-23-45/.hydra/overrides.yaml b/outputs/2026-06-30/19-23-45/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..efdd985de2643ea464ba05109c047aedfa2b3a80 --- /dev/null +++ b/outputs/2026-06-30/19-23-45/.hydra/overrides.yaml @@ -0,0 +1,61 @@ +- +name=train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=false +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=1000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.0 +- algorithm.memory_noise.dynamic_max_fraction=0.25 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=10 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=150000 diff --git a/outputs/2026-06-30/19-23-45/main.log b/outputs/2026-06-30/19-23-45/main.log new file mode 100644 index 0000000000000000000000000000000000000000..fb118e548e7a958d9c864d0be3994cbe8c993c72 --- /dev/null +++ b/outputs/2026-06-30/19-23-45/main.log @@ -0,0 +1,144 @@ +[2026-06-30 19:24:35,855][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-06-30 19:24:35,856][pytorch_lightning.utilities.rank_zero][INFO] - Missing checkpoint keys: timestamp_embedding.mlp.0.weight, timestamp_embedding.mlp.0.bias, timestamp_embedding.mlp.2.weight, timestamp_embedding.mlp.2.bias, spatial_rotary_emb.freqs, temporal_rotary_emb.freqs, pose_embedder.model.0.weight, pose_embedder.model.0.bias, ... (+560 more) +[2026-06-30 19:24:35,949][pytorch_lightning.utilities.rank_zero][INFO] - Zeroed frame-memory gate slices for base checkpoint load: blocks.0.r_adaLN_modulation.1.weight, blocks.0.r_adaLN_modulation.1.bias, blocks.1.r_adaLN_modulation.1.weight, blocks.1.r_adaLN_modulation.1.bias, blocks.2.r_adaLN_modulation.1.weight, blocks.2.r_adaLN_modulation.1.bias, blocks.3.r_adaLN_modulation.1.weight, blocks.3.r_adaLN_modulation.1.bias, ... (+24 more) +[2026-06-30 19:24:35,950][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: 302 tensors. +[2026-06-30 19:24:36,037][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-06-30 19:24:36,075][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: 228 tensors. +[2026-06-30 19:24:36,120][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-30 19:24:36,269][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-30 19:24:36,270][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-30 19:24:36,270][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-30 19:24:36,270][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-30 19:24:36,270][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-30 19:25:10,900][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-30 19:25:10,900][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-30 19:25:11,204][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-30 19:25:11,273][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-30 19:25:17,651][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-30 19:25:23,015][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:23,015][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:23,015][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:23,015][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:24,713][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-30 19:25:25,768][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 19:25:25,769][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 19:25:25,769][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 19:25:25,769][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 20:08:00,782][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,367][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,368][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,368][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,368][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,368][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,368][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,368][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,369][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,369][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,369][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,369][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,369][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,369][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,369][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,369][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,369][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,369][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,370][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,370][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,370][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,370][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,370][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:08:03,370][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-01 06:33:03,044][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=30000` reached. +[2026-07-01 06:33:24,811][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-01 06:33:24,812][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:33:24,812][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:33:24,812][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:33:24,813][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:33:25,110][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-01 06:33:25,110][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-01 06:33:25,110][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-01 06:33:25,110][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-01 06:33:25,111][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-01 06:35:15,557][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/callbacks/model_checkpoint.py:639: Checkpoint directory /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/checkpoints exists and is not empty. + +[2026-07-01 06:35:15,558][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/checkpoints/curriculum_01_no_updown_step30000.ckpt +[2026-07-01 06:35:15,572][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:35:15,573][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:35:15,575][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:35:15,582][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:35:34,844][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:35:34,850][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:35:34,852][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:35:34,855][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:35:35,707][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-07-01 06:35:36,832][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/checkpoints/curriculum_01_no_updown_step30000.ckpt +[2026-07-01 06:35:36,897][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:35:36,897][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:35:36,897][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:35:36,897][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:35:38,048][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/loops/training_epoch_loop.py:154: You're resuming from a checkpoint that ended before the epoch ended. This can cause unreliable results if further training is done. Consider using an end-of-epoch checkpoint + +[2026-07-01 23:51:08,642][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-07-01 23:51:08,642][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-07-01 23:51:08,643][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-01 23:51:08,643][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-01 23:51:08,646][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-07-01 23:51:08,646][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-01 23:51:08,647][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-07-01 23:51:08,647][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-30/19-23-46/.hydra/config.yaml b/outputs/2026-06-30/19-23-46/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..44622f84726ec574fca73aa589ee22826a9d62a2 --- /dev/null +++ b/outputs/2026-06-30/19-23-46/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 150000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 10 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 30000 + dataset: + wo_updown: true + - name: with_updown + until_step: 150000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: false + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.0 + dynamic_max_fraction: 0.25 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-30/19-23-46/.hydra/hydra.yaml b/outputs/2026-06-30/19-23-46/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3aa259464e34b44a8d149c297e9fe78a4896bb0c --- /dev/null +++ b/outputs/2026-06-30/19-23-46/.hydra/hydra.yaml @@ -0,0 +1,219 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=false + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=1000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.0 + - algorithm.memory_noise.dynamic_max_fraction=0.25 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=10 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=150000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.0,algorithm.memory_noise.dynamic_max_fraction=0.25,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=1000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.causal=false,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=10,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]},experiment.training.max_steps=150000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-30/19-23-46 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-30/19-23-46/.hydra/overrides.yaml b/outputs/2026-06-30/19-23-46/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..efdd985de2643ea464ba05109c047aedfa2b3a80 --- /dev/null +++ b/outputs/2026-06-30/19-23-46/.hydra/overrides.yaml @@ -0,0 +1,61 @@ +- +name=train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_noncausal_a2d4r2_route_ntok8_8h200_bs16_global128_150k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=false +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=1000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.0 +- algorithm.memory_noise.dynamic_max_fraction=0.25 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=10 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=150000 diff --git a/outputs/2026-06-30/19-23-46/main.log b/outputs/2026-06-30/19-23-46/main.log new file mode 100644 index 0000000000000000000000000000000000000000..86a073a2a85061ee06a23e8f2503bd9e9b1b697b --- /dev/null +++ b/outputs/2026-06-30/19-23-46/main.log @@ -0,0 +1,11 @@ +[2026-06-30 19:25:10,904][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-30 19:25:23,015][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:25,769][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:33:24,812][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:35:15,583][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:35:34,844][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:35:36,897][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 23:51:08,649][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-07-01 23:51:08,650][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-06-30/19-23-51/.hydra/config.yaml b/outputs/2026-06-30/19-23-51/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8ff5d963c0848b9bb2ea4f52dadab7f3ccf650b6 --- /dev/null +++ b/outputs/2026-06-30/19-23-51/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 150000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 10 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 30000 + dataset: + wo_updown: true + - name: with_updown + until_step: 150000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.0 + dynamic_max_fraction: 0.25 + revisit_max_fraction: 0.25 + noise_route: + anchor: all + dynamic: all + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-30/19-23-51/.hydra/hydra.yaml b/outputs/2026-06-30/19-23-51/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7d5f60ec77d8f901fece75bd729afe31cf35aae9 --- /dev/null +++ b/outputs/2026-06-30/19-23-51/.hydra/hydra.yaml @@ -0,0 +1,219 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=1000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=all + - algorithm.noise_route.dynamic=all + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.0 + - algorithm.memory_noise.dynamic_max_fraction=0.25 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=10 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=150000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.0,algorithm.memory_noise.dynamic_max_fraction=0.25,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=all,algorithm.noise_route.dynamic=all,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=1000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=10,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]},experiment.training.max_steps=150000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-30/19-23-51 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-30/19-23-51/.hydra/overrides.yaml b/outputs/2026-06-30/19-23-51/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7e3f2353d03287af5ccbd829323aa8d3239cc6e1 --- /dev/null +++ b/outputs/2026-06-30/19-23-51/.hydra/overrides.yaml @@ -0,0 +1,61 @@ +- +name=train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=1000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=all +- algorithm.noise_route.dynamic=all +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.0 +- algorithm.memory_noise.dynamic_max_fraction=0.25 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=10 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=150000 diff --git a/outputs/2026-06-30/19-23-51/main.log b/outputs/2026-06-30/19-23-51/main.log new file mode 100644 index 0000000000000000000000000000000000000000..505ec37d440805746ae02faca39eef77f8e31917 --- /dev/null +++ b/outputs/2026-06-30/19-23-51/main.log @@ -0,0 +1,54 @@ +[2026-06-30 19:25:17,363][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-06-30 19:25:17,364][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-06-30 19:25:17,366][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-06-30 19:25:17,366][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-06-30 19:25:17,367][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-06-30 19:25:17,368][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-06-30 19:25:29,276][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:29,276][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:29,276][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:29,276][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:29,276][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:29,276][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:32,437][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 19:25:32,437][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 19:25:32,437][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 19:25:32,437][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 19:25:32,437][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 19:25:32,437][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:56:56,994][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:56:56,994][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:56:56,994][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:56:56,994][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:56:56,994][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:56:56,994][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:58:41,444][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:58:41,448][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:58:41,449][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:58:41,452][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:58:41,456][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:58:41,457][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:59:01,169][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:59:01,171][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:59:01,176][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:59:01,190][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:59:01,191][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:59:01,192][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:59:03,260][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:59:03,260][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:59:03,260][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:59:03,260][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:59:03,260][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:59:03,260][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-06-30/19-23-52/.hydra/config.yaml b/outputs/2026-06-30/19-23-52/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8ff5d963c0848b9bb2ea4f52dadab7f3ccf650b6 --- /dev/null +++ b/outputs/2026-06-30/19-23-52/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 150000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 10 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 30000 + dataset: + wo_updown: true + - name: with_updown + until_step: 150000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.0 + dynamic_max_fraction: 0.25 + revisit_max_fraction: 0.25 + noise_route: + anchor: all + dynamic: all + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-06-30/19-23-52/.hydra/hydra.yaml b/outputs/2026-06-30/19-23-52/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..6f047fd30b78cc1da0300e3837b19fec0ad404e7 --- /dev/null +++ b/outputs/2026-06-30/19-23-52/.hydra/hydra.yaml @@ -0,0 +1,219 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=1000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=all + - algorithm.noise_route.dynamic=all + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.0 + - algorithm.memory_noise.dynamic_max_fraction=0.25 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=10 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=150000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.0,algorithm.memory_noise.dynamic_max_fraction=0.25,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=all,algorithm.noise_route.dynamic=all,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=1000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=10,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]},experiment.training.max_steps=150000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-30/19-23-52 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-06-30/19-23-52/.hydra/overrides.yaml b/outputs/2026-06-30/19-23-52/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7e3f2353d03287af5ccbd829323aa8d3239cc6e1 --- /dev/null +++ b/outputs/2026-06-30/19-23-52/.hydra/overrides.yaml @@ -0,0 +1,61 @@ +- +name=train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=1000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=all +- algorithm.noise_route.dynamic=all +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.0 +- algorithm.memory_noise.dynamic_max_fraction=0.25 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:150000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=10 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=150000 diff --git a/outputs/2026-06-30/19-23-52/main.log b/outputs/2026-06-30/19-23-52/main.log new file mode 100644 index 0000000000000000000000000000000000000000..7034a9f7955637cca654c1f04012053c267d497d --- /dev/null +++ b/outputs/2026-06-30/19-23-52/main.log @@ -0,0 +1,119 @@ +[2026-06-30 19:24:40,658][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-06-30 19:24:40,659][pytorch_lightning.utilities.rank_zero][INFO] - Missing checkpoint keys: timestamp_embedding.mlp.0.weight, timestamp_embedding.mlp.0.bias, timestamp_embedding.mlp.2.weight, timestamp_embedding.mlp.2.bias, spatial_rotary_emb.freqs, temporal_rotary_emb.freqs, pose_embedder.model.0.weight, pose_embedder.model.0.bias, ... (+560 more) +[2026-06-30 19:24:40,754][pytorch_lightning.utilities.rank_zero][INFO] - Zeroed frame-memory gate slices for base checkpoint load: blocks.0.r_adaLN_modulation.1.weight, blocks.0.r_adaLN_modulation.1.bias, blocks.1.r_adaLN_modulation.1.weight, blocks.1.r_adaLN_modulation.1.bias, blocks.2.r_adaLN_modulation.1.weight, blocks.2.r_adaLN_modulation.1.bias, blocks.3.r_adaLN_modulation.1.weight, blocks.3.r_adaLN_modulation.1.bias, ... (+24 more) +[2026-06-30 19:24:40,754][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: 302 tensors. +[2026-06-30 19:24:40,838][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-06-30 19:24:40,876][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: 228 tensors. +[2026-06-30 19:24:40,920][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-06-30 19:24:41,066][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-06-30 19:24:41,066][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-06-30 19:24:41,066][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-06-30 19:24:41,066][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-06-30 19:24:41,067][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-06-30 19:25:17,365][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-06-30 19:25:17,740][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-06-30 19:25:24,027][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-06-30 19:25:29,276][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:29,276][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-06-30 19:25:32,435][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-06-30 19:25:32,436][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 19:25:32,437][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-06-30 20:09:28,129][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,297][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,297][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,297][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,298][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,298][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,298][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,298][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,298][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,298][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,299][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,299][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,299][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,299][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,299][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,299][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,299][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,300][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,300][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,300][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,300][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,300][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,300][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-06-30 20:09:43,300][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-01 06:56:34,634][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=30000` reached. +[2026-07-01 06:56:56,994][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-01 06:56:56,994][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:56:56,995][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-01 06:56:58,603][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-01 06:56:58,603][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-01 06:56:58,604][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-01 06:56:58,604][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-01 06:56:58,605][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-01 06:58:41,430][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/callbacks/model_checkpoint.py:639: Checkpoint directory /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k/checkpoints exists and is not empty. + +[2026-07-01 06:58:41,431][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k/checkpoints/curriculum_01_no_updown_step30000.ckpt +[2026-07-01 06:58:41,445][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:58:41,454][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-01 06:59:01,193][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:59:01,245][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-01 06:59:02,086][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-07-01 06:59:03,240][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_no_route_anchorclean_a2d4r2_ntok8_8h200_bs16_global128_150k/checkpoints/curriculum_01_no_updown_step30000.ckpt +[2026-07-01 06:59:03,260][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:59:03,260][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-01 06:59:04,353][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/loops/training_epoch_loop.py:154: You're resuming from a checkpoint that ended before the epoch ended. This can cause unreliable results if further training is done. Consider using an end-of-epoch checkpoint + +[2026-07-03 07:28:34,306][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=150000` reached. diff --git a/outputs/2026-07-03/18-31-12/.hydra/config.yaml b/outputs/2026-07-03/18-31-12/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4ba66d1cd911a8035a8bc6462ae3b82c497859b1 --- /dev/null +++ b/outputs/2026-07-03/18-31-12/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 120000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 30000 + dataset: + wo_updown: true + - name: with_updown + until_step: 120000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.0 + dynamic_max_fraction: 0.0 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-03/18-31-12/.hydra/hydra.yaml b/outputs/2026-07-03/18-31-12/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9548a87efcee3f7aab8dd96a51cb21db59d32309 --- /dev/null +++ b/outputs/2026-07-03/18-31-12/.hydra/hydra.yaml @@ -0,0 +1,219 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=0 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=1000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.0 + - algorithm.memory_noise.dynamic_max_fraction=0.0 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=120000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.0,algorithm.memory_noise.dynamic_max_fraction=0.0,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=1000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=0,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]},experiment.training.max_steps=120000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-03/18-31-12 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-03/18-31-12/.hydra/overrides.yaml b/outputs/2026-07-03/18-31-12/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a11b1f8dea651b89d720494b95235887b8322312 --- /dev/null +++ b/outputs/2026-07-03/18-31-12/.hydra/overrides.yaml @@ -0,0 +1,61 @@ +- +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=0 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=1000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.0 +- algorithm.memory_noise.dynamic_max_fraction=0.0 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=120000 diff --git a/outputs/2026-07-03/18-31-12/main.log b/outputs/2026-07-03/18-31-12/main.log new file mode 100644 index 0000000000000000000000000000000000000000..8f126fd377588eebac9a8f0a692ca4a3eee90eb9 --- /dev/null +++ b/outputs/2026-07-03/18-31-12/main.log @@ -0,0 +1,9 @@ +[2026-07-03 18:33:09,225][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-07-03 18:33:20,962][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-03 18:33:23,499][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-04 03:08:28,447][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-04 03:10:25,149][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-04 03:10:45,413][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-04 03:10:47,312][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-07-03/18-31-13/.hydra/config.yaml b/outputs/2026-07-03/18-31-13/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4ba66d1cd911a8035a8bc6462ae3b82c497859b1 --- /dev/null +++ b/outputs/2026-07-03/18-31-13/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 120000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 30000 + dataset: + wo_updown: true + - name: with_updown + until_step: 120000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.0 + dynamic_max_fraction: 0.0 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-03/18-31-13/.hydra/hydra.yaml b/outputs/2026-07-03/18-31-13/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..6a9a01a4185288d7bf52556d3ede1f6956f901da --- /dev/null +++ b/outputs/2026-07-03/18-31-13/.hydra/hydra.yaml @@ -0,0 +1,219 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=0 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=1000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.0 + - algorithm.memory_noise.dynamic_max_fraction=0.0 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=120000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.0,algorithm.memory_noise.dynamic_max_fraction=0.0,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=1000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=0,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]},experiment.training.max_steps=120000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-03/18-31-13 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-03/18-31-13/.hydra/overrides.yaml b/outputs/2026-07-03/18-31-13/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a11b1f8dea651b89d720494b95235887b8322312 --- /dev/null +++ b/outputs/2026-07-03/18-31-13/.hydra/overrides.yaml @@ -0,0 +1,61 @@ +- +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=0 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=1000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.0 +- algorithm.memory_noise.dynamic_max_fraction=0.0 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=120000 diff --git a/outputs/2026-07-03/18-31-13/main.log b/outputs/2026-07-03/18-31-13/main.log new file mode 100644 index 0000000000000000000000000000000000000000..259a67716c02623e545a2c318b845d957577fbc0 --- /dev/null +++ b/outputs/2026-07-03/18-31-13/main.log @@ -0,0 +1,9 @@ +[2026-07-03 18:33:09,226][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-07-03 18:33:20,962][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-03 18:33:23,499][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-04 03:08:28,448][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-04 03:10:25,144][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-04 03:10:45,412][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-04 03:10:47,312][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-07-03/18-31-16/.hydra/config.yaml b/outputs/2026-07-03/18-31-16/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4ba66d1cd911a8035a8bc6462ae3b82c497859b1 --- /dev/null +++ b/outputs/2026-07-03/18-31-16/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 120000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 30000 + dataset: + wo_updown: true + - name: with_updown + until_step: 120000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.0 + dynamic_max_fraction: 0.0 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-03/18-31-16/.hydra/hydra.yaml b/outputs/2026-07-03/18-31-16/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c59c7d19cb0e6413c55d3caaee7b12ab96bdd96d --- /dev/null +++ b/outputs/2026-07-03/18-31-16/.hydra/hydra.yaml @@ -0,0 +1,219 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=0 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=1000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.0 + - algorithm.memory_noise.dynamic_max_fraction=0.0 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=120000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.0,algorithm.memory_noise.dynamic_max_fraction=0.0,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=1000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=0,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]},experiment.training.max_steps=120000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-03/18-31-16 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-03/18-31-16/.hydra/overrides.yaml b/outputs/2026-07-03/18-31-16/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a11b1f8dea651b89d720494b95235887b8322312 --- /dev/null +++ b/outputs/2026-07-03/18-31-16/.hydra/overrides.yaml @@ -0,0 +1,61 @@ +- +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=0 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=1000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.0 +- algorithm.memory_noise.dynamic_max_fraction=0.0 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=120000 diff --git a/outputs/2026-07-03/18-31-16/main.log b/outputs/2026-07-03/18-31-16/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e498b55af1d3a408c410704a800afee09bf5b5ad --- /dev/null +++ b/outputs/2026-07-03/18-31-16/main.log @@ -0,0 +1,18 @@ +[2026-07-03 18:33:09,228][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-07-03 18:33:09,492][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-07-03 18:33:20,962][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-03 18:33:20,962][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-03 18:33:23,499][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-03 18:33:23,499][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-04 03:08:28,447][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-04 03:08:28,447][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-04 03:10:25,142][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-04 03:10:25,147][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-04 03:10:45,414][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-04 03:10:45,418][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-04 03:10:47,312][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-04 03:10:47,312][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-07-03/18-31-17/.hydra/config.yaml b/outputs/2026-07-03/18-31-17/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4ba66d1cd911a8035a8bc6462ae3b82c497859b1 --- /dev/null +++ b/outputs/2026-07-03/18-31-17/.hydra/config.yaml @@ -0,0 +1,214 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 120000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 30000 + dataset: + wo_updown: true + - name: with_updown + until_step: 120000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 0 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + dynamic: + selection_policy: event_triggered + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + pose_preselect_topk: 64 + candidate_chunk_size: 64 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.0 + dynamic_max_fraction: 0.0 + revisit_max_fraction: 0.25 + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + add_timestamp_embedding: true + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-03/18-31-17/.hydra/hydra.yaml b/outputs/2026-07-03/18-31-17/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..81c727304641d159a19a1061efb6bd71ff155ace --- /dev/null +++ b/outputs/2026-07-03/18-31-17/.hydra/hydra.yaml @@ -0,0 +1,219 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=0 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.pose_similarity_threshold=0.6 + - dataset.memory_selection.training_use_plucker=true + - dataset.memory_selection.training_plucker_weight=1.0 + - dataset.memory_selection.fov_overlap_threshold=0.6 + - dataset.memory_selection.min_total_selected_coverage=0.1 + - dataset.memory_selection.local_context_exclusion_frames=8 + - dataset.memory_selection.anchor_diverse_selection=true + - +dataset.memory_selection.pose_preselect_topk=64 + - +dataset.memory_selection.candidate_chunk_size=64 + - dataset.memory_selection.dynamic.selection_policy=event_triggered + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.chunk_size=1 + - algorithm.warmup_steps=1000 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=true + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.0 + - algorithm.memory_noise.dynamic_max_fraction=0.0 + - algorithm.memory_noise.revisit_max_fraction=0.25 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=120000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+dataset.memory_selection.candidate_chunk_size=64,+dataset.memory_selection.pose_preselect_topk=64,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_noise.anchor_max_fraction=0.0,algorithm.memory_noise.dynamic_max_fraction=0.0,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.25,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=1000,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.anchor_diverse_selection=true,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.selection_policy=event_triggered,dataset.memory_selection.enabled=true,dataset.memory_selection.fov_overlap_threshold=0.6,dataset.memory_selection.local_context_exclusion_frames=8,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=0,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.min_total_selected_coverage=0.1,dataset.memory_selection.pose_similarity_threshold=0.6,dataset.memory_selection.training_plucker_weight=1.0,dataset.memory_selection.training_use_plucker=true,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]},experiment.training.max_steps=120000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-03/18-31-17 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-03/18-31-17/.hydra/overrides.yaml b/outputs/2026-07-03/18-31-17/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a11b1f8dea651b89d720494b95235887b8322312 --- /dev/null +++ b/outputs/2026-07-03/18-31-17/.hydra/overrides.yaml @@ -0,0 +1,61 @@ +- +name=train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=0 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.pose_similarity_threshold=0.6 +- dataset.memory_selection.training_use_plucker=true +- dataset.memory_selection.training_plucker_weight=1.0 +- dataset.memory_selection.fov_overlap_threshold=0.6 +- dataset.memory_selection.min_total_selected_coverage=0.1 +- dataset.memory_selection.local_context_exclusion_frames=8 +- dataset.memory_selection.anchor_diverse_selection=true +- +dataset.memory_selection.pose_preselect_topk=64 +- +dataset.memory_selection.candidate_chunk_size=64 +- dataset.memory_selection.dynamic.selection_policy=event_triggered +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.chunk_size=1 +- algorithm.warmup_steps=1000 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=true +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.0 +- algorithm.memory_noise.dynamic_max_fraction=0.0 +- algorithm.memory_noise.revisit_max_fraction=0.25 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:30000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=120000 diff --git a/outputs/2026-07-03/18-31-17/main.log b/outputs/2026-07-03/18-31-17/main.log new file mode 100644 index 0000000000000000000000000000000000000000..f3466edeed9d6a0fc56459c3d3772a484d05646a --- /dev/null +++ b/outputs/2026-07-03/18-31-17/main.log @@ -0,0 +1,137 @@ +[2026-07-03 18:32:25,228][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-07-03 18:32:25,229][pytorch_lightning.utilities.rank_zero][INFO] - Missing checkpoint keys: timestamp_embedding.mlp.0.weight, timestamp_embedding.mlp.0.bias, timestamp_embedding.mlp.2.weight, timestamp_embedding.mlp.2.bias, spatial_rotary_emb.freqs, temporal_rotary_emb.freqs, pose_embedder.model.0.weight, pose_embedder.model.0.bias, ... (+560 more) +[2026-07-03 18:32:25,318][pytorch_lightning.utilities.rank_zero][INFO] - Zeroed frame-memory gate slices for base checkpoint load: blocks.0.r_adaLN_modulation.1.weight, blocks.0.r_adaLN_modulation.1.bias, blocks.1.r_adaLN_modulation.1.weight, blocks.1.r_adaLN_modulation.1.bias, blocks.2.r_adaLN_modulation.1.weight, blocks.2.r_adaLN_modulation.1.bias, blocks.3.r_adaLN_modulation.1.weight, blocks.3.r_adaLN_modulation.1.bias, ... (+24 more) +[2026-07-03 18:32:25,319][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: 302 tensors. +[2026-07-03 18:32:25,386][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-07-03 18:32:25,417][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: 228 tensors. +[2026-07-03 18:32:25,458][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-03 18:32:25,688][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-03 18:32:25,688][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-03 18:32:25,688][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-03 18:32:25,688][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-03 18:32:25,688][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-03 18:33:09,225][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-07-03 18:33:09,227][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-07-03 18:33:09,228][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-07-03 18:33:09,850][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-07-03 18:33:15,942][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-07-03 18:33:20,962][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-03 18:33:20,962][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-03 18:33:20,962][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-03 18:33:20,962][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-03 18:33:22,391][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-07-03 18:33:23,499][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-03 18:33:23,499][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-03 18:33:23,499][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-03 18:33:23,499][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-03 19:06:03,246][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,990][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,990][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,990][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,990][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,991][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,991][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,991][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,991][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,991][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,991][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,991][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,991][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,991][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,991][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,992][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,992][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,992][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,992][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,992][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,992][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,992][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,992][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-03 19:06:07,992][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-04 03:08:07,035][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=30000` reached. +[2026-07-04 03:08:28,447][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-04 03:08:28,447][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-04 03:08:28,447][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-04 03:08:28,447][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-04 03:08:28,448][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-04 03:08:29,459][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-04 03:08:29,460][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-04 03:08:29,460][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-04 03:08:29,460][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-04 03:08:29,461][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-04 03:10:25,131][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/callbacks/model_checkpoint.py:639: Checkpoint directory /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/checkpoints exists and is not empty. + +[2026-07-04 03:10:25,132][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/checkpoints/curriculum_01_no_updown_step30000.ckpt +[2026-07-04 03:10:25,139][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-04 03:10:25,142][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-04 03:10:25,151][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-04 03:10:25,151][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-04 03:10:45,416][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-04 03:10:45,417][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-04 03:10:45,419][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-04 03:10:45,419][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-04 03:10:46,226][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +438 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,112.927 Total estimated model params size (MB) +[2026-07-04 03:10:47,297][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_event_trigger_anchorclean_causal_a2d0r2_route_ntok8_8h200_bs16_global128_120k/checkpoints/curriculum_01_no_updown_step30000.ckpt +[2026-07-04 03:10:47,312][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-04 03:10:47,312][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-04 03:10:47,312][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-04 03:10:47,312][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-04 03:10:48,296][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/loops/training_epoch_loop.py:154: You're resuming from a checkpoint that ended before the epoch ended. This can cause unreliable results if further training is done. Consider using an end-of-epoch checkpoint + +[2026-07-05 06:27:45,563][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=120000` reached. diff --git a/outputs/2026-07-06/21-35-50/.hydra/config.yaml b/outputs/2026-07-06/21-35-50/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..024cbcfd2e5d40b89d32122ae2898bbb2bfdd2f4 --- /dev/null +++ b/outputs/2026-07-06/21-35-50/.hydra/config.yaml @@ -0,0 +1,218 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 120000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 120000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: fov_greedy + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.5 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + memory_attention_key_only_geometry: true + add_timestamp_embedding: false + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-06/21-35-50/.hydra/hydra.yaml b/outputs/2026-07-06/21-35-50/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f854ccc81128776a515d6162d63fb82263def609 --- /dev/null +++ b/outputs/2026-07-06/21-35-50/.hydra/hydra.yaml @@ -0,0 +1,212 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=fov_greedy + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.memory_attention_key_only_geometry=true + - algorithm.add_timestamp_embedding=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_max_fraction=0.50 + - algorithm.memory_noise.revisit_max_fraction=0.50 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=120000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.add_timestamp_embedding=false,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_max_fraction=0.50,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.50,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=fov_greedy,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]},experiment.training.max_steps=120000,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-06/21-35-50 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-06/21-35-50/.hydra/overrides.yaml b/outputs/2026-07-06/21-35-50/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b30d277b6349296325e79fe83f661bca8f3765a3 --- /dev/null +++ b/outputs/2026-07-06/21-35-50/.hydra/overrides.yaml @@ -0,0 +1,54 @@ +- +name=train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=fov_greedy +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.memory_attention_key_only_geometry=true +- algorithm.add_timestamp_embedding=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_max_fraction=0.50 +- algorithm.memory_noise.revisit_max_fraction=0.50 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=120000 diff --git a/outputs/2026-07-06/21-35-50/main.log b/outputs/2026-07-06/21-35-50/main.log new file mode 100644 index 0000000000000000000000000000000000000000..3facc78b7788c9efde43f0113837d9057646cf59 --- /dev/null +++ b/outputs/2026-07-06/21-35-50/main.log @@ -0,0 +1,172 @@ +[2026-07-06 21:36:46,882][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-07-06 21:36:46,883][pytorch_lightning.utilities.rank_zero][INFO] - Missing checkpoint keys: spatial_rotary_emb.freqs, temporal_rotary_emb.freqs, pose_embedder.model.0.weight, pose_embedder.model.0.bias, pose_embedder.model.2.weight, pose_embedder.model.2.bias, blocks.0.r_mlp.fc1.weight, blocks.0.r_mlp.fc1.bias, ... (+556 more) +[2026-07-06 21:36:46,973][pytorch_lightning.utilities.rank_zero][INFO] - Zeroed frame-memory gate slices for base checkpoint load: blocks.0.r_adaLN_modulation.1.weight, blocks.0.r_adaLN_modulation.1.bias, blocks.1.r_adaLN_modulation.1.weight, blocks.1.r_adaLN_modulation.1.bias, blocks.2.r_adaLN_modulation.1.weight, blocks.2.r_adaLN_modulation.1.bias, blocks.3.r_adaLN_modulation.1.weight, blocks.3.r_adaLN_modulation.1.bias, ... (+24 more) +[2026-07-06 21:36:46,974][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: 302 tensors. +[2026-07-06 21:36:47,074][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-07-06 21:36:47,106][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: 228 tensors. +[2026-07-06 21:36:47,167][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-06 21:36:47,375][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-06 21:36:47,375][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-06 21:36:47,375][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-06 21:36:47,375][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-06 21:36:47,376][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-06 21:37:17,625][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-07-06 21:37:17,625][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-07-06 21:37:17,625][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-07-06 21:37:17,628][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-07-06 21:37:17,630][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-07-06 21:37:17,637][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-07-06 21:37:17,789][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-07-06 21:37:17,828][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-07-06 21:37:23,966][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-07-06 21:37:29,128][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-06 21:37:29,128][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-06 21:37:29,128][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-06 21:37:29,128][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-06 21:37:29,128][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-06 21:37:29,128][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-06 21:37:29,128][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-06 21:37:29,128][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-06 21:37:30,871][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +437 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,107.676 Total estimated model params size (MB) +[2026-07-06 21:37:33,445][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-06 21:37:33,445][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-06 21:37:33,445][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-06 21:37:33,445][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-06 21:37:33,445][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-06 21:37:33,445][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-06 21:37:33,445][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-06 21:37:33,445][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-06 22:18:24,382][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,168][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,168][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,168][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,168][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,168][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,168][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,169][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,169][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,169][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,169][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,169][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,169][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,169][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,169][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,169][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,169][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,170][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,170][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,170][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,170][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,170][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,170][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:18:26,170][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-07 11:56:24,926][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=40000` reached. +[2026-07-07 11:56:46,313][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-07 11:56:46,313][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-07 11:56:46,313][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-07 11:56:46,313][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-07 11:56:46,313][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-07 11:56:46,313][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-07 11:56:46,313][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-07 11:56:46,314][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-07 11:56:46,314][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-07 11:56:47,376][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-07 11:56:47,376][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-07 11:56:47,376][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-07 11:56:47,376][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-07 11:56:47,377][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-07 11:58:28,388][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/callbacks/model_checkpoint.py:639: Checkpoint directory /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k/checkpoints exists and is not empty. + +[2026-07-07 11:58:28,390][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k/checkpoints/curriculum_01_no_updown_step40000.ckpt +[2026-07-07 11:58:28,396][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-07 11:58:28,396][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-07 11:58:28,402][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-07 11:58:28,403][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-07 11:58:28,405][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-07 11:58:28,407][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-07 11:58:28,410][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-07 11:58:28,412][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-07 11:58:51,356][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-07 11:58:51,356][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-07 11:58:51,358][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-07 11:58:51,359][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-07 11:58:51,359][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-07 11:58:51,360][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-07 11:58:51,364][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-07 11:58:51,365][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-07 11:58:52,200][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +437 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,107.676 Total estimated model params size (MB) +[2026-07-07 11:58:53,323][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k/checkpoints/curriculum_01_no_updown_step40000.ckpt +[2026-07-07 11:58:53,349][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-07 11:58:53,349][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-07 11:58:53,350][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-07 11:58:53,350][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-07 11:58:53,350][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-07 11:58:53,349][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-07 11:58:53,350][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-07 11:58:53,350][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-07 11:58:54,450][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/loops/training_epoch_loop.py:154: You're resuming from a checkpoint that ended before the epoch ended. This can cause unreliable results if further training is done. Consider using an end-of-epoch checkpoint + diff --git a/outputs/2026-07-06/21-51-21/.hydra/config.yaml b/outputs/2026-07-06/21-51-21/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..da455977ca073cd59b672c4c743d3c545a8898f6 --- /dev/null +++ b/outputs/2026-07-06/21-51-21/.hydra/config.yaml @@ -0,0 +1,225 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 120000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 120000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: fov_greedy + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: 500 + memory_noise: + enabled: true + mode: source_conditioned + anchor_max_fraction: 0.25 + dynamic_clean_max_fraction: 0.25 + revisit_clean_max_fraction: 0.25 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_source_embeddings: true + memory_attention_key_only_geometry: true + add_timestamp_embedding: false + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-06/21-51-21/.hydra/hydra.yaml b/outputs/2026-07-06/21-51-21/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..eb355f507f4e625c592a41498bbeb4b1823b7420 --- /dev/null +++ b/outputs/2026-07-06/21-51-21/.hydra/hydra.yaml @@ -0,0 +1,223 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - experiment.num_nodes=1 + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=fov_greedy + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_source_embeddings=true + - algorithm.memory_attention_key_only_geometry=true + - algorithm.add_timestamp_embedding=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_conditioned + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_clean_max_fraction=0.25 + - algorithm.memory_noise.revisit_clean_max_fraction=0.25 + - algorithm.memory_noise.dynamic_generated_max_fraction=0.5 + - algorithm.memory_noise.revisit_generated_max_fraction=0.5 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.validation.inference_mode=false + - experiment.validation.val_every_n_step=2000 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.training.max_steps=120000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.add_timestamp_embedding=false,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_clean_max_fraction=0.25,algorithm.memory_noise.dynamic_generated_max_fraction=0.5,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_conditioned,algorithm.memory_noise.revisit_clean_max_fraction=0.25,algorithm.memory_noise.revisit_generated_max_fraction=0.5,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_source_embeddings=true,algorithm.memory_source_simulation.training_generated_suffix_max_frames=500,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.use_memory_attention=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=fov_greedy,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]},experiment.training.max_steps=120000,experiment.validation.batch_size=1,experiment.validation.inference_mode=false,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-06/21-51-21 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-06/21-51-21/.hydra/overrides.yaml b/outputs/2026-07-06/21-51-21/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f64ce23a93228172d6a76c751896c3e0f6141614 --- /dev/null +++ b/outputs/2026-07-06/21-51-21/.hydra/overrides.yaml @@ -0,0 +1,65 @@ +- +name=train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- experiment.num_nodes=1 +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=fov_greedy +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_source_embeddings=true +- algorithm.memory_attention_key_only_geometry=true +- algorithm.add_timestamp_embedding=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_conditioned +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_clean_max_fraction=0.25 +- algorithm.memory_noise.revisit_clean_max_fraction=0.25 +- algorithm.memory_noise.dynamic_generated_max_fraction=0.5 +- algorithm.memory_noise.revisit_generated_max_fraction=0.5 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.validation.inference_mode=false +- experiment.validation.val_every_n_step=2000 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.training.max_steps=120000 diff --git a/outputs/2026-07-06/21-51-21/main.log b/outputs/2026-07-06/21-51-21/main.log new file mode 100644 index 0000000000000000000000000000000000000000..30febcb0a2e0dca0e43ea64beda86b69e90152b7 --- /dev/null +++ b/outputs/2026-07-06/21-51-21/main.log @@ -0,0 +1,117 @@ +[2026-07-06 21:52:26,797][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-07-06 21:52:26,798][pytorch_lightning.utilities.rank_zero][INFO] - Missing checkpoint keys: spatial_rotary_emb.freqs, temporal_rotary_emb.freqs, pose_embedder.model.0.weight, pose_embedder.model.0.bias, pose_embedder.model.2.weight, pose_embedder.model.2.bias, blocks.0.r_mlp.fc1.weight, blocks.0.r_mlp.fc1.bias, ... (+300 more) +[2026-07-06 21:52:26,886][pytorch_lightning.utilities.rank_zero][INFO] - Zeroed frame-memory gate slices for base checkpoint load: blocks.0.r_adaLN_modulation.1.weight, blocks.0.r_adaLN_modulation.1.bias, blocks.1.r_adaLN_modulation.1.weight, blocks.1.r_adaLN_modulation.1.bias, blocks.2.r_adaLN_modulation.1.weight, blocks.2.r_adaLN_modulation.1.bias, blocks.3.r_adaLN_modulation.1.weight, blocks.3.r_adaLN_modulation.1.bias, ... (+24 more) +[2026-07-06 21:52:26,887][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: 302 tensors. +[2026-07-06 21:52:26,966][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-07-06 21:52:26,998][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: 228 tensors. +[2026-07-06 21:52:27,038][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-06 21:52:27,185][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-06 21:52:27,185][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-06 21:52:27,185][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-06 21:52:27,185][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-06 21:52:27,186][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-06 21:53:04,974][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-07-06 21:53:04,978][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-07-06 21:53:04,980][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-07-06 21:53:10,508][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-07-06 21:53:16,355][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-06 21:53:16,355][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-06 21:53:16,355][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-06 21:53:17,196][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +302 M Trainable params +839 M Non-trainable params +1.1 B Total params +4,569.297 Total estimated model params size (MB) +[2026-07-06 21:53:19,136][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-06 21:53:19,135][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-06 21:53:19,136][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-06 22:31:44,044][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,195][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,195][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,196][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,196][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,196][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,196][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,196][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,196][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,196][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,196][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,196][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,197][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,197][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,197][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,197][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,197][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,197][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-06 22:31:50,197][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-07 11:23:58,728][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=40000` reached. +[2026-07-07 11:24:18,527][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-07 11:24:18,527][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-07 11:24:18,527][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-07 11:24:18,528][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-07 11:24:19,875][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-07 11:24:19,875][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-07 11:24:19,875][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-07 11:24:19,875][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-07 11:24:19,876][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-07 11:26:08,907][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/callbacks/model_checkpoint.py:639: Checkpoint directory /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/checkpoints exists and is not empty. + +[2026-07-07 11:26:08,908][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/checkpoints/curriculum_01_no_updown_step40000.ckpt +[2026-07-07 11:26:08,914][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-07 11:26:08,915][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-07 11:26:08,925][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-07 11:26:27,138][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-07 11:26:27,147][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-07 11:26:27,152][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-07 11:26:27,963][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +302 M Trainable params +839 M Non-trainable params +1.1 B Total params +4,569.297 Total estimated model params size (MB) +[2026-07-07 11:26:28,925][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/checkpoints/curriculum_01_no_updown_step40000.ckpt +[2026-07-07 11:26:28,942][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-07 11:26:28,942][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-07 11:26:28,942][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-07 11:26:29,819][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/loops/training_epoch_loop.py:154: You're resuming from a checkpoint that ended before the epoch ended. This can cause unreliable results if further training is done. Consider using an end-of-epoch checkpoint + diff --git a/outputs/2026-07-06/21-51-22/.hydra/config.yaml b/outputs/2026-07-06/21-51-22/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..da455977ca073cd59b672c4c743d3c545a8898f6 --- /dev/null +++ b/outputs/2026-07-06/21-51-22/.hydra/config.yaml @@ -0,0 +1,225 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 120000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 120000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: fov_greedy + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: 500 + memory_noise: + enabled: true + mode: source_conditioned + anchor_max_fraction: 0.25 + dynamic_clean_max_fraction: 0.25 + revisit_clean_max_fraction: 0.25 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_source_embeddings: true + memory_attention_key_only_geometry: true + add_timestamp_embedding: false + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-06/21-51-22/.hydra/hydra.yaml b/outputs/2026-07-06/21-51-22/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..6ebcaef60392bb4cf77f28af6f5810653d7fde13 --- /dev/null +++ b/outputs/2026-07-06/21-51-22/.hydra/hydra.yaml @@ -0,0 +1,223 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - experiment.num_nodes=1 + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=fov_greedy + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_source_embeddings=true + - algorithm.memory_attention_key_only_geometry=true + - algorithm.add_timestamp_embedding=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_conditioned + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_clean_max_fraction=0.25 + - algorithm.memory_noise.revisit_clean_max_fraction=0.25 + - algorithm.memory_noise.dynamic_generated_max_fraction=0.5 + - algorithm.memory_noise.revisit_generated_max_fraction=0.5 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.validation.inference_mode=false + - experiment.validation.val_every_n_step=2000 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.training.max_steps=120000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.add_timestamp_embedding=false,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_clean_max_fraction=0.25,algorithm.memory_noise.dynamic_generated_max_fraction=0.5,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_conditioned,algorithm.memory_noise.revisit_clean_max_fraction=0.25,algorithm.memory_noise.revisit_generated_max_fraction=0.5,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_source_embeddings=true,algorithm.memory_source_simulation.training_generated_suffix_max_frames=500,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.use_memory_attention=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=fov_greedy,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]},experiment.training.max_steps=120000,experiment.validation.batch_size=1,experiment.validation.inference_mode=false,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-06/21-51-22 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-06/21-51-22/.hydra/overrides.yaml b/outputs/2026-07-06/21-51-22/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f64ce23a93228172d6a76c751896c3e0f6141614 --- /dev/null +++ b/outputs/2026-07-06/21-51-22/.hydra/overrides.yaml @@ -0,0 +1,65 @@ +- +name=train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- experiment.num_nodes=1 +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=fov_greedy +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_source_embeddings=true +- algorithm.memory_attention_key_only_geometry=true +- algorithm.add_timestamp_embedding=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_conditioned +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_clean_max_fraction=0.25 +- algorithm.memory_noise.revisit_clean_max_fraction=0.25 +- algorithm.memory_noise.dynamic_generated_max_fraction=0.5 +- algorithm.memory_noise.revisit_generated_max_fraction=0.5 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.validation.inference_mode=false +- experiment.validation.val_every_n_step=2000 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.training.max_steps=120000 diff --git a/outputs/2026-07-06/21-51-22/main.log b/outputs/2026-07-06/21-51-22/main.log new file mode 100644 index 0000000000000000000000000000000000000000..23d3aa6bf2a72545d2dbe3aeac0af89bd7fa2d34 --- /dev/null +++ b/outputs/2026-07-06/21-51-22/main.log @@ -0,0 +1,36 @@ +[2026-07-06 21:53:04,975][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-07-06 21:53:04,976][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-07-06 21:53:04,978][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-07-06 21:53:04,983][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-07-06 21:53:16,355][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-06 21:53:16,355][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-06 21:53:16,355][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-06 21:53:16,355][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-06 21:53:19,136][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-06 21:53:19,136][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-06 21:53:19,136][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-06 21:53:19,136][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-07 11:24:18,527][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-07 11:24:18,527][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-07 11:24:18,527][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-07 11:24:18,527][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-07 11:26:08,916][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-07 11:26:08,917][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-07 11:26:08,925][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-07 11:26:08,926][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-07 11:26:27,137][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-07 11:26:27,141][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-07 11:26:27,147][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-07 11:26:27,152][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-07 11:26:28,942][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-07 11:26:28,942][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-07 11:26:28,942][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-07 11:26:28,942][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-07-06/21-51-23/.hydra/config.yaml b/outputs/2026-07-06/21-51-23/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..da455977ca073cd59b672c4c743d3c545a8898f6 --- /dev/null +++ b/outputs/2026-07-06/21-51-23/.hydra/config.yaml @@ -0,0 +1,225 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 120000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 120000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: fov_greedy + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: 500 + memory_noise: + enabled: true + mode: source_conditioned + anchor_max_fraction: 0.25 + dynamic_clean_max_fraction: 0.25 + revisit_clean_max_fraction: 0.25 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_source_embeddings: true + memory_attention_key_only_geometry: true + add_timestamp_embedding: false + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-06/21-51-23/.hydra/hydra.yaml b/outputs/2026-07-06/21-51-23/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..57bbe4f3dcef779e5d3ff1bd6e495aae7ad26e5a --- /dev/null +++ b/outputs/2026-07-06/21-51-23/.hydra/hydra.yaml @@ -0,0 +1,223 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - experiment.num_nodes=1 + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=fov_greedy + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_source_embeddings=true + - algorithm.memory_attention_key_only_geometry=true + - algorithm.add_timestamp_embedding=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_conditioned + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_clean_max_fraction=0.25 + - algorithm.memory_noise.revisit_clean_max_fraction=0.25 + - algorithm.memory_noise.dynamic_generated_max_fraction=0.5 + - algorithm.memory_noise.revisit_generated_max_fraction=0.5 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.validation.inference_mode=false + - experiment.validation.val_every_n_step=2000 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.training.max_steps=120000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.add_timestamp_embedding=false,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_clean_max_fraction=0.25,algorithm.memory_noise.dynamic_generated_max_fraction=0.5,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_conditioned,algorithm.memory_noise.revisit_clean_max_fraction=0.25,algorithm.memory_noise.revisit_generated_max_fraction=0.5,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_source_embeddings=true,algorithm.memory_source_simulation.training_generated_suffix_max_frames=500,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.use_memory_attention=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=fov_greedy,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]},experiment.training.max_steps=120000,experiment.validation.batch_size=1,experiment.validation.inference_mode=false,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-06/21-51-23 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-06/21-51-23/.hydra/overrides.yaml b/outputs/2026-07-06/21-51-23/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f64ce23a93228172d6a76c751896c3e0f6141614 --- /dev/null +++ b/outputs/2026-07-06/21-51-23/.hydra/overrides.yaml @@ -0,0 +1,65 @@ +- +name=train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- experiment.num_nodes=1 +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=fov_greedy +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_source_embeddings=true +- algorithm.memory_attention_key_only_geometry=true +- algorithm.add_timestamp_embedding=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_conditioned +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_clean_max_fraction=0.25 +- algorithm.memory_noise.revisit_clean_max_fraction=0.25 +- algorithm.memory_noise.dynamic_generated_max_fraction=0.5 +- algorithm.memory_noise.revisit_generated_max_fraction=0.5 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.validation.inference_mode=false +- experiment.validation.val_every_n_step=2000 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.training.max_steps=120000 diff --git a/outputs/2026-07-06/21-51-23/main.log b/outputs/2026-07-06/21-51-23/main.log new file mode 100644 index 0000000000000000000000000000000000000000..ce90917957fbba4b7d9e9da8d44efdef4cff4393 --- /dev/null +++ b/outputs/2026-07-06/21-51-23/main.log @@ -0,0 +1,9 @@ +[2026-07-06 21:53:04,976][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-07-06 21:53:16,355][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-06 21:53:19,136][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-07 11:24:18,527][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-07 11:26:08,926][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-07 11:26:27,281][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-07 11:26:28,942][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-07-10/11-07-57/.hydra/config.yaml b/outputs/2026-07-10/11-07-57/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..024cbcfd2e5d40b89d32122ae2898bbb2bfdd2f4 --- /dev/null +++ b/outputs/2026-07-10/11-07-57/.hydra/config.yaml @@ -0,0 +1,218 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 120000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 120000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: fov_greedy + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_noise: + enabled: true + mode: random_cleaner_fraction + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.5 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + memory_attention_key_only_geometry: true + add_timestamp_embedding: false + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-10/11-07-57/.hydra/hydra.yaml b/outputs/2026-07-10/11-07-57/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a83f2de971ec639f20fe45736434e1d35ad1c200 --- /dev/null +++ b/outputs/2026-07-10/11-07-57/.hydra/hydra.yaml @@ -0,0 +1,212 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=fov_greedy + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.memory_attention_key_only_geometry=true + - algorithm.add_timestamp_embedding=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=random_cleaner_fraction + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_max_fraction=0.50 + - algorithm.memory_noise.revisit_max_fraction=0.50 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.validation.val_every_n_step=2000 + - experiment.training.max_steps=120000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.add_timestamp_embedding=false,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_max_fraction=0.50,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=random_cleaner_fraction,algorithm.memory_noise.revisit_max_fraction=0.50,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=fov_greedy,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]},experiment.training.max_steps=120000,experiment.validation.batch_size=1,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-10/11-07-57 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-10/11-07-57/.hydra/overrides.yaml b/outputs/2026-07-10/11-07-57/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b30d277b6349296325e79fe83f661bca8f3765a3 --- /dev/null +++ b/outputs/2026-07-10/11-07-57/.hydra/overrides.yaml @@ -0,0 +1,54 @@ +- +name=train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=fov_greedy +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.memory_attention_key_only_geometry=true +- algorithm.add_timestamp_embedding=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=random_cleaner_fraction +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_max_fraction=0.50 +- algorithm.memory_noise.revisit_max_fraction=0.50 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.validation.val_every_n_step=2000 +- experiment.training.max_steps=120000 diff --git a/outputs/2026-07-10/11-07-57/main.log b/outputs/2026-07-10/11-07-57/main.log new file mode 100644 index 0000000000000000000000000000000000000000..633d067f4db8ed71f3cea2f863d66f1d04d6fd02 --- /dev/null +++ b/outputs/2026-07-10/11-07-57/main.log @@ -0,0 +1,97 @@ +[2026-07-10 11:08:59,514][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-10 11:08:59,689][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-10 11:08:59,690][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-10 11:08:59,690][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-10 11:08:59,690][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-10 11:08:59,690][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-10 11:10:51,417][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-07-10 11:10:51,427][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-07-10 11:10:51,433][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-07-10 11:10:51,465][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-07-10 11:10:51,487][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-07-10 11:10:51,506][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-07-10 11:10:51,550][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-07-10 11:10:51,740][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-07-10 11:10:57,124][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-07-10 11:11:02,391][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k/checkpoints/epoch2_step112000.ckpt +[2026-07-10 11:11:25,397][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-10 11:11:25,400][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-10 11:11:25,402][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-10 11:11:25,402][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-10 11:11:25,407][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-10 11:11:25,411][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-10 11:11:25,413][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-10 11:11:25,455][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-10 11:11:26,386][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 1.0 B +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +437 M Trainable params +839 M Non-trainable params +1.3 B Total params +5,107.676 Total estimated model params size (MB) +[2026-07-10 11:11:27,399][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_curric_multiview_causal_a2d4r2_route_ntok8_8h200_bs16_global128_120k/checkpoints/epoch2_step112000.ckpt +[2026-07-10 11:11:30,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-10 11:11:30,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-10 11:11:30,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-10 11:11:30,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-10 11:11:30,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-10 11:11:30,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-10 11:11:30,775][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-10 11:11:30,774][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-10 11:22:02,333][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,930][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,930][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,930][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,930][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,930][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,931][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,931][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,931][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,931][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,931][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,931][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,931][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,932][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,932][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,932][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,932][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,932][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,932][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,932][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,932][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,933][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,933][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:17,933][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/generation_length_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 14:19:05,436][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=120000` reached. diff --git a/outputs/2026-07-10/11-09-21/.hydra/config.yaml b/outputs/2026-07-10/11-09-21/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..da455977ca073cd59b672c4c743d3c545a8898f6 --- /dev/null +++ b/outputs/2026-07-10/11-09-21/.hydra/config.yaml @@ -0,0 +1,225 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 120000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 120000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: fov_greedy + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: 500 + memory_noise: + enabled: true + mode: source_conditioned + anchor_max_fraction: 0.25 + dynamic_clean_max_fraction: 0.25 + revisit_clean_max_fraction: 0.25 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_source_embeddings: true + memory_attention_key_only_geometry: true + add_timestamp_embedding: false + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-10/11-09-21/.hydra/hydra.yaml b/outputs/2026-07-10/11-09-21/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f4c62c559e870de09228adbce5db6f8bda31248c --- /dev/null +++ b/outputs/2026-07-10/11-09-21/.hydra/hydra.yaml @@ -0,0 +1,223 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - experiment.num_nodes=1 + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=fov_greedy + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_source_embeddings=true + - algorithm.memory_attention_key_only_geometry=true + - algorithm.add_timestamp_embedding=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_conditioned + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_clean_max_fraction=0.25 + - algorithm.memory_noise.revisit_clean_max_fraction=0.25 + - algorithm.memory_noise.dynamic_generated_max_fraction=0.5 + - algorithm.memory_noise.revisit_generated_max_fraction=0.5 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.validation.inference_mode=false + - experiment.validation.val_every_n_step=2000 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.training.max_steps=120000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.add_timestamp_embedding=false,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_clean_max_fraction=0.25,algorithm.memory_noise.dynamic_generated_max_fraction=0.5,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_conditioned,algorithm.memory_noise.revisit_clean_max_fraction=0.25,algorithm.memory_noise.revisit_generated_max_fraction=0.5,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_source_embeddings=true,algorithm.memory_source_simulation.training_generated_suffix_max_frames=500,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.use_memory_attention=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=fov_greedy,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]},experiment.training.max_steps=120000,experiment.validation.batch_size=1,experiment.validation.inference_mode=false,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-10/11-09-21 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-10/11-09-21/.hydra/overrides.yaml b/outputs/2026-07-10/11-09-21/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f64ce23a93228172d6a76c751896c3e0f6141614 --- /dev/null +++ b/outputs/2026-07-10/11-09-21/.hydra/overrides.yaml @@ -0,0 +1,65 @@ +- +name=train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- experiment.num_nodes=1 +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=fov_greedy +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_source_embeddings=true +- algorithm.memory_attention_key_only_geometry=true +- algorithm.add_timestamp_embedding=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_conditioned +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_clean_max_fraction=0.25 +- algorithm.memory_noise.revisit_clean_max_fraction=0.25 +- algorithm.memory_noise.dynamic_generated_max_fraction=0.5 +- algorithm.memory_noise.revisit_generated_max_fraction=0.5 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.validation.inference_mode=false +- experiment.validation.val_every_n_step=2000 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.training.max_steps=120000 diff --git a/outputs/2026-07-10/11-09-21/main.log b/outputs/2026-07-10/11-09-21/main.log new file mode 100644 index 0000000000000000000000000000000000000000..4f6630a17d9f4e35fc7bca15c0638f78531832b4 --- /dev/null +++ b/outputs/2026-07-10/11-09-21/main.log @@ -0,0 +1,3 @@ +[2026-07-10 11:12:23,342][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-07-10 11:12:55,460][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-10 11:13:00,647][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-07-10/11-09-22/.hydra/config.yaml b/outputs/2026-07-10/11-09-22/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..da455977ca073cd59b672c4c743d3c545a8898f6 --- /dev/null +++ b/outputs/2026-07-10/11-09-22/.hydra/config.yaml @@ -0,0 +1,225 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 120000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 120000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + pose_similarity_threshold: 0.6 + training_use_plucker: true + training_plucker_weight: 1.0 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: fov_greedy + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: 500 + memory_noise: + enabled: true + mode: source_conditioned + anchor_max_fraction: 0.25 + dynamic_clean_max_fraction: 0.25 + revisit_clean_max_fraction: 0.25 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_source_embeddings: true + memory_attention_key_only_geometry: true + add_timestamp_embedding: false + ref_mode: sequential + focal_length: 0.35 + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-10/11-09-22/.hydra/hydra.yaml b/outputs/2026-07-10/11-09-22/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9274707308dbee6001cbe4592e990144e37b6e29 --- /dev/null +++ b/outputs/2026-07-10/11-09-22/.hydra/hydra.yaml @@ -0,0 +1,223 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - experiment.num_nodes=1 + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=fov_greedy + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_source_embeddings=true + - algorithm.memory_attention_key_only_geometry=true + - algorithm.add_timestamp_embedding=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_conditioned + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_clean_max_fraction=0.25 + - algorithm.memory_noise.revisit_clean_max_fraction=0.25 + - algorithm.memory_noise.dynamic_generated_max_fraction=0.5 + - algorithm.memory_noise.revisit_generated_max_fraction=0.5 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.validation.inference_mode=false + - experiment.validation.val_every_n_step=2000 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.training.max_steps=120000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.add_timestamp_embedding=false,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_clean_max_fraction=0.25,algorithm.memory_noise.dynamic_generated_max_fraction=0.5,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_conditioned,algorithm.memory_noise.revisit_clean_max_fraction=0.25,algorithm.memory_noise.revisit_generated_max_fraction=0.5,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_source_embeddings=true,algorithm.memory_source_simulation.training_generated_suffix_max_frames=500,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.use_memory_attention=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=fov_greedy,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]},experiment.training.max_steps=120000,experiment.validation.batch_size=1,experiment.validation.inference_mode=false,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-10/11-09-22 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-10/11-09-22/.hydra/overrides.yaml b/outputs/2026-07-10/11-09-22/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f64ce23a93228172d6a76c751896c3e0f6141614 --- /dev/null +++ b/outputs/2026-07-10/11-09-22/.hydra/overrides.yaml @@ -0,0 +1,65 @@ +- +name=train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- experiment.num_nodes=1 +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=fov_greedy +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_source_embeddings=true +- algorithm.memory_attention_key_only_geometry=true +- algorithm.add_timestamp_embedding=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_conditioned +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_clean_max_fraction=0.25 +- algorithm.memory_noise.revisit_clean_max_fraction=0.25 +- algorithm.memory_noise.dynamic_generated_max_fraction=0.5 +- algorithm.memory_noise.revisit_generated_max_fraction=0.5 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:120000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.validation.inference_mode=false +- experiment.validation.val_every_n_step=2000 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.training.max_steps=120000 diff --git a/outputs/2026-07-10/11-09-22/main.log b/outputs/2026-07-10/11-09-22/main.log new file mode 100644 index 0000000000000000000000000000000000000000..5af77c1d8461cf44d6d55e713b6a6f863808c401 --- /dev/null +++ b/outputs/2026-07-10/11-09-22/main.log @@ -0,0 +1,84 @@ +[2026-07-10 11:10:27,557][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-10 11:10:27,706][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-10 11:10:27,706][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-10 11:10:27,707][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-10 11:10:27,707][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-10 11:10:27,707][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-10 11:12:23,287][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-07-10 11:12:23,288][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-07-10 11:12:23,292][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-07-10 11:12:23,316][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-07-10 11:12:23,363][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-07-10 11:12:23,383][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-07-10 11:12:23,624][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-07-10 11:12:30,003][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-07-10 11:12:35,202][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/checkpoints/epoch2_step118000.ckpt +[2026-07-10 11:12:55,466][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-10 11:12:55,529][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-10 11:12:55,536][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-10 11:12:55,542][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-10 11:12:55,544][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-10 11:12:55,551][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-10 11:12:55,600][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-10 11:12:58,876][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +302 M Trainable params +839 M Non-trainable params +1.1 B Total params +4,569.297 Total estimated model params size (MB) +[2026-07-10 11:12:59,797][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_base_8h200_bs16_global128_120k/checkpoints/epoch2_step118000.ckpt +[2026-07-10 11:13:00,646][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-10 11:13:00,647][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-10 11:13:00,647][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-10 11:13:00,647][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-10 11:13:00,647][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-10 11:13:00,647][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-10 11:13:00,647][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-10 11:22:38,659][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,910][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,910][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,910][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,910][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,910][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,911][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,911][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,911][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,911][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,911][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,911][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,911][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,911][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,911][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,911][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,912][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,912][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 11:22:38,912][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-10 12:04:25,679][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=120000` reached. diff --git a/outputs/2026-07-14/20-11-44/.hydra/config.yaml b/outputs/2026-07-14/20-11-44/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..69764c8f248b29d03fca0ef1651010ccba69d4f4 --- /dev/null +++ b/outputs/2026-07-14/20-11-44/.hydra/config.yaml @@ -0,0 +1,228 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: 500 + memory_noise: + enabled: true + mode: source_conditioned + anchor_max_fraction: 0.25 + dynamic_clean_max_fraction: 0.25 + revisit_clean_max_fraction: 0.25 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_source_embeddings: true + memory_attention_key_only_geometry: true + add_timestamp_embedding: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-14/20-11-44/.hydra/hydra.yaml b/outputs/2026-07-14/20-11-44/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..002ebaf408a077e8ba585bcdd0e33941153ae2bf --- /dev/null +++ b/outputs/2026-07-14/20-11-44/.hydra/hydra.yaml @@ -0,0 +1,223 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - experiment.num_nodes=1 + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_source_embeddings=true + - algorithm.memory_attention_key_only_geometry=true + - algorithm.add_timestamp_embedding=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_conditioned + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_clean_max_fraction=0.25 + - algorithm.memory_noise.revisit_clean_max_fraction=0.25 + - algorithm.memory_noise.dynamic_generated_max_fraction=0.5 + - algorithm.memory_noise.revisit_generated_max_fraction=0.5 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.validation.inference_mode=false + - experiment.validation.val_every_n_step=2000 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.add_timestamp_embedding=false,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_clean_max_fraction=0.25,algorithm.memory_noise.dynamic_generated_max_fraction=0.5,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_conditioned,algorithm.memory_noise.revisit_clean_max_fraction=0.25,algorithm.memory_noise.revisit_generated_max_fraction=0.5,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_source_embeddings=true,algorithm.memory_source_simulation.training_generated_suffix_max_frames=500,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.use_memory_attention=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.validation.batch_size=1,experiment.validation.inference_mode=false,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-14/20-11-44 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-14/20-11-44/.hydra/overrides.yaml b/outputs/2026-07-14/20-11-44/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..65d50ff618ec1345f166c53bf26330c31a08ad9c --- /dev/null +++ b/outputs/2026-07-14/20-11-44/.hydra/overrides.yaml @@ -0,0 +1,65 @@ +- +name=train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- experiment.num_nodes=1 +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_source_embeddings=true +- algorithm.memory_attention_key_only_geometry=true +- algorithm.add_timestamp_embedding=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_conditioned +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_clean_max_fraction=0.25 +- algorithm.memory_noise.revisit_clean_max_fraction=0.25 +- algorithm.memory_noise.dynamic_generated_max_fraction=0.5 +- algorithm.memory_noise.revisit_generated_max_fraction=0.5 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.validation.inference_mode=false +- experiment.validation.val_every_n_step=2000 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-07-14/20-11-44/main.log b/outputs/2026-07-14/20-11-44/main.log new file mode 100644 index 0000000000000000000000000000000000000000..4112bc37fffd38256bfb2e4d87af02f3787c8a88 --- /dev/null +++ b/outputs/2026-07-14/20-11-44/main.log @@ -0,0 +1,22 @@ +[2026-07-14 20:13:39,379][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-07-14 20:13:39,379][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-07-14 20:13:52,879][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-14 20:13:52,879][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-14 20:13:56,511][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-14 20:13:56,511][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-14 23:24:39,472][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-14 23:24:39,472][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-14 23:25:08,407][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-14 23:25:08,411][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-14 23:25:19,458][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-14 23:25:19,466][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-14 23:25:20,505][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-14 23:25:20,505][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 10:08:09,962][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 10:08:10,138][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 10:08:12,611][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 10:08:12,611][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-07-14/20-11-47/.hydra/config.yaml b/outputs/2026-07-14/20-11-47/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..69764c8f248b29d03fca0ef1651010ccba69d4f4 --- /dev/null +++ b/outputs/2026-07-14/20-11-47/.hydra/config.yaml @@ -0,0 +1,228 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: 500 + memory_noise: + enabled: true + mode: source_conditioned + anchor_max_fraction: 0.25 + dynamic_clean_max_fraction: 0.25 + revisit_clean_max_fraction: 0.25 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_source_embeddings: true + memory_attention_key_only_geometry: true + add_timestamp_embedding: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-14/20-11-47/.hydra/hydra.yaml b/outputs/2026-07-14/20-11-47/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..2741a803347bc24d9bff3dcd48d5c901b5511eee --- /dev/null +++ b/outputs/2026-07-14/20-11-47/.hydra/hydra.yaml @@ -0,0 +1,223 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - experiment.num_nodes=1 + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_source_embeddings=true + - algorithm.memory_attention_key_only_geometry=true + - algorithm.add_timestamp_embedding=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_conditioned + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_clean_max_fraction=0.25 + - algorithm.memory_noise.revisit_clean_max_fraction=0.25 + - algorithm.memory_noise.dynamic_generated_max_fraction=0.5 + - algorithm.memory_noise.revisit_generated_max_fraction=0.5 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.validation.inference_mode=false + - experiment.validation.val_every_n_step=2000 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.add_timestamp_embedding=false,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_clean_max_fraction=0.25,algorithm.memory_noise.dynamic_generated_max_fraction=0.5,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_conditioned,algorithm.memory_noise.revisit_clean_max_fraction=0.25,algorithm.memory_noise.revisit_generated_max_fraction=0.5,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_source_embeddings=true,algorithm.memory_source_simulation.training_generated_suffix_max_frames=500,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.use_memory_attention=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.validation.batch_size=1,experiment.validation.inference_mode=false,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-14/20-11-47 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-14/20-11-47/.hydra/overrides.yaml b/outputs/2026-07-14/20-11-47/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..65d50ff618ec1345f166c53bf26330c31a08ad9c --- /dev/null +++ b/outputs/2026-07-14/20-11-47/.hydra/overrides.yaml @@ -0,0 +1,65 @@ +- +name=train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- experiment.num_nodes=1 +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_source_embeddings=true +- algorithm.memory_attention_key_only_geometry=true +- algorithm.add_timestamp_embedding=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_conditioned +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_clean_max_fraction=0.25 +- algorithm.memory_noise.revisit_clean_max_fraction=0.25 +- algorithm.memory_noise.dynamic_generated_max_fraction=0.5 +- algorithm.memory_noise.revisit_generated_max_fraction=0.5 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.validation.inference_mode=false +- experiment.validation.val_every_n_step=2000 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-07-14/20-11-47/main.log b/outputs/2026-07-14/20-11-47/main.log new file mode 100644 index 0000000000000000000000000000000000000000..6a6570457aaab58efb542ef140621f9dd5228196 --- /dev/null +++ b/outputs/2026-07-14/20-11-47/main.log @@ -0,0 +1,11 @@ +[2026-07-14 20:13:39,376][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-07-14 20:13:52,879][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-14 20:13:56,511][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-14 23:24:39,471][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-14 23:25:08,294][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-14 23:25:19,463][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-14 23:25:20,505][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 10:08:09,849][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 10:08:12,611][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-07-14/20-11-48/.hydra/config.yaml b/outputs/2026-07-14/20-11-48/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..69764c8f248b29d03fca0ef1651010ccba69d4f4 --- /dev/null +++ b/outputs/2026-07-14/20-11-48/.hydra/config.yaml @@ -0,0 +1,228 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: 500 + memory_noise: + enabled: true + mode: source_conditioned + anchor_max_fraction: 0.25 + dynamic_clean_max_fraction: 0.25 + revisit_clean_max_fraction: 0.25 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_source_embeddings: true + memory_attention_key_only_geometry: true + add_timestamp_embedding: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-14/20-11-48/.hydra/hydra.yaml b/outputs/2026-07-14/20-11-48/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..2e5fbfeab1fac15e2dfd15166303ab01e002da43 --- /dev/null +++ b/outputs/2026-07-14/20-11-48/.hydra/hydra.yaml @@ -0,0 +1,223 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - experiment.num_nodes=1 + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_source_embeddings=true + - algorithm.memory_attention_key_only_geometry=true + - algorithm.add_timestamp_embedding=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_conditioned + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_clean_max_fraction=0.25 + - algorithm.memory_noise.revisit_clean_max_fraction=0.25 + - algorithm.memory_noise.dynamic_generated_max_fraction=0.5 + - algorithm.memory_noise.revisit_generated_max_fraction=0.5 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.validation.inference_mode=false + - experiment.validation.val_every_n_step=2000 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.add_timestamp_embedding=false,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_clean_max_fraction=0.25,algorithm.memory_noise.dynamic_generated_max_fraction=0.5,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_conditioned,algorithm.memory_noise.revisit_clean_max_fraction=0.25,algorithm.memory_noise.revisit_generated_max_fraction=0.5,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_source_embeddings=true,algorithm.memory_source_simulation.training_generated_suffix_max_frames=500,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.use_memory_attention=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.validation.batch_size=1,experiment.validation.inference_mode=false,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-14/20-11-48 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-14/20-11-48/.hydra/overrides.yaml b/outputs/2026-07-14/20-11-48/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..65d50ff618ec1345f166c53bf26330c31a08ad9c --- /dev/null +++ b/outputs/2026-07-14/20-11-48/.hydra/overrides.yaml @@ -0,0 +1,65 @@ +- +name=train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- experiment.num_nodes=1 +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_source_embeddings=true +- algorithm.memory_attention_key_only_geometry=true +- algorithm.add_timestamp_embedding=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_conditioned +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_clean_max_fraction=0.25 +- algorithm.memory_noise.revisit_clean_max_fraction=0.25 +- algorithm.memory_noise.dynamic_generated_max_fraction=0.5 +- algorithm.memory_noise.revisit_generated_max_fraction=0.5 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.validation.inference_mode=false +- experiment.validation.val_every_n_step=2000 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-07-14/20-11-48/main.log b/outputs/2026-07-14/20-11-48/main.log new file mode 100644 index 0000000000000000000000000000000000000000..2aa3c9f5869f1a2573768375a9c1996cd1fbc604 --- /dev/null +++ b/outputs/2026-07-14/20-11-48/main.log @@ -0,0 +1,155 @@ +[2026-07-14 20:12:54,296][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-07-14 20:12:54,296][pytorch_lightning.utilities.rank_zero][INFO] - Missing checkpoint keys: spatial_rotary_emb.freqs, temporal_rotary_emb.freqs, pose_embedder.model.0.weight, pose_embedder.model.0.bias, pose_embedder.model.2.weight, pose_embedder.model.2.bias, blocks.0.r_mlp.fc1.weight, blocks.0.r_mlp.fc1.bias, ... (+300 more) +[2026-07-14 20:12:54,386][pytorch_lightning.utilities.rank_zero][INFO] - Zeroed frame-memory gate slices for base checkpoint load: blocks.0.r_adaLN_modulation.1.weight, blocks.0.r_adaLN_modulation.1.bias, blocks.1.r_adaLN_modulation.1.weight, blocks.1.r_adaLN_modulation.1.bias, blocks.2.r_adaLN_modulation.1.weight, blocks.2.r_adaLN_modulation.1.bias, blocks.3.r_adaLN_modulation.1.weight, blocks.3.r_adaLN_modulation.1.bias, ... (+24 more) +[2026-07-14 20:12:54,387][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: 302 tensors. +[2026-07-14 20:12:54,472][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-07-14 20:12:54,510][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: 228 tensors. +[2026-07-14 20:12:54,567][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-14 20:12:54,772][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-14 20:12:54,773][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-14 20:12:54,773][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-14 20:12:54,773][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-14 20:12:54,773][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-14 20:13:39,377][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-07-14 20:13:39,381][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-07-14 20:13:39,381][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-07-14 20:13:40,063][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-07-14 20:13:47,346][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-07-14 20:13:52,879][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-14 20:13:52,879][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-14 20:13:52,879][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-14 20:13:52,879][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-14 20:13:56,509][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +302 M Trainable params +839 M Non-trainable params +1.1 B Total params +4,569.297 Total estimated model params size (MB) +[2026-07-14 20:13:56,511][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-14 20:13:56,510][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-14 20:13:56,511][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-14 20:13:56,511][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-14 20:51:57,962][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,596][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,596][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,596][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,596][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,596][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,596][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,597][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,597][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,597][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,597][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,597][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,597][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,597][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,597][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,597][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,597][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,598][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 20:52:11,598][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-14 23:24:27,800][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=10000` reached. +[2026-07-14 23:24:39,471][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-14 23:24:39,471][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-14 23:24:39,472][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-14 23:24:39,472][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-14 23:24:39,472][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-14 23:24:39,909][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-14 23:24:39,909][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-14 23:24:39,909][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-14 23:24:39,909][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-14 23:24:39,910][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-14 23:25:08,177][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/callbacks/model_checkpoint.py:639: Checkpoint directory /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/checkpoints exists and is not empty. + +[2026-07-14 23:25:08,285][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_memory_only_step10000.ckpt +[2026-07-14 23:25:08,294][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-14 23:25:08,295][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-14 23:25:08,401][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-14 23:25:08,409][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-14 23:25:19,462][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-14 23:25:19,463][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-14 23:25:19,464][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-14 23:25:19,464][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-14 23:25:20,166][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +910 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,569.297 Total estimated model params size (MB) +[2026-07-14 23:25:20,501][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/checkpoints/curriculum_01_no_updown_memory_only_step10000.ckpt +[2026-07-14 23:25:20,505][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-14 23:25:20,505][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-14 23:25:20,505][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-14 23:25:20,505][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-14 23:25:21,003][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/loops/training_epoch_loop.py:154: You're resuming from a checkpoint that ended before the epoch ended. This can cause unreliable results if further training is done. Consider using an end-of-epoch checkpoint + +[2026-07-15 10:05:27,767][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=40000` reached. +[2026-07-15 10:05:51,297][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-15 10:05:53,388][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-15 10:05:53,388][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-15 10:05:53,388][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-15 10:05:53,388][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-15 10:05:53,388][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-15 10:07:49,987][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/checkpoints/curriculum_02_no_updown_step40000.ckpt +[2026-07-15 10:08:09,946][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 10:08:10,000][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 10:08:10,073][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 10:08:10,163][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 10:08:10,891][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +910 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,569.297 Total estimated model params size (MB) +[2026-07-15 10:08:11,917][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/checkpoints/curriculum_02_no_updown_step40000.ckpt +[2026-07-15 10:08:12,611][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 10:08:12,611][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 10:08:12,611][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 10:08:12,611][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 18:46:56,515][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=200000` reached. diff --git a/outputs/2026-07-14/20-11-49/.hydra/config.yaml b/outputs/2026-07-14/20-11-49/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..69764c8f248b29d03fca0ef1651010ccba69d4f4 --- /dev/null +++ b/outputs/2026-07-14/20-11-49/.hydra/config.yaml @@ -0,0 +1,228 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: 500 + memory_noise: + enabled: true + mode: source_conditioned + anchor_max_fraction: 0.25 + dynamic_clean_max_fraction: 0.25 + revisit_clean_max_fraction: 0.25 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_source_embeddings: true + memory_attention_key_only_geometry: true + add_timestamp_embedding: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +name: train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/ +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-14/20-11-49/.hydra/hydra.yaml b/outputs/2026-07-14/20-11-49/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..43d7f89278c0dc73ff8afab1d7aa9530b8c33cc8 --- /dev/null +++ b/outputs/2026-07-14/20-11-49/.hydra/hydra.yaml @@ -0,0 +1,223 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/ + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - experiment.num_nodes=1 + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_source_embeddings=true + - algorithm.memory_attention_key_only_geometry=true + - algorithm.add_timestamp_embedding=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_conditioned + - algorithm.memory_noise.anchor_max_fraction=0.25 + - algorithm.memory_noise.dynamic_clean_max_fraction=0.25 + - algorithm.memory_noise.revisit_clean_max_fraction=0.25 + - algorithm.memory_noise.dynamic_generated_max_fraction=0.5 + - algorithm.memory_noise.revisit_generated_max_fraction=0.5 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.validation.inference_mode=false + - experiment.validation.val_every_n_step=2000 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.add_timestamp_embedding=false,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.anchor_max_fraction=0.25,algorithm.memory_noise.dynamic_clean_max_fraction=0.25,algorithm.memory_noise.dynamic_generated_max_fraction=0.5,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_conditioned,algorithm.memory_noise.revisit_clean_max_fraction=0.25,algorithm.memory_noise.revisit_generated_max_fraction=0.5,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_source_embeddings=true,algorithm.memory_source_simulation.training_generated_suffix_max_frames=500,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.use_memory_attention=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.validation.batch_size=1,experiment.validation.inference_mode=false,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-14/20-11-49 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-14/20-11-49/.hydra/overrides.yaml b/outputs/2026-07-14/20-11-49/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..65d50ff618ec1345f166c53bf26330c31a08ad9c --- /dev/null +++ b/outputs/2026-07-14/20-11-49/.hydra/overrides.yaml @@ -0,0 +1,65 @@ +- +name=train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/ +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- experiment.num_nodes=1 +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_source_embeddings=true +- algorithm.memory_attention_key_only_geometry=true +- algorithm.add_timestamp_embedding=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=500 +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_conditioned +- algorithm.memory_noise.anchor_max_fraction=0.25 +- algorithm.memory_noise.dynamic_clean_max_fraction=0.25 +- algorithm.memory_noise.revisit_clean_max_fraction=0.25 +- algorithm.memory_noise.dynamic_generated_max_fraction=0.5 +- algorithm.memory_noise.revisit_generated_max_fraction=0.5 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.validation.inference_mode=false +- experiment.validation.val_every_n_step=2000 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-07-14/20-11-49/main.log b/outputs/2026-07-14/20-11-49/main.log new file mode 100644 index 0000000000000000000000000000000000000000..5cefc86d7b239f5e50e5dd724d204bb86e1deec6 --- /dev/null +++ b/outputs/2026-07-14/20-11-49/main.log @@ -0,0 +1,11 @@ +[2026-07-14 20:13:39,375][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-07-14 20:13:52,879][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-14 20:13:56,511][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-14 23:24:39,471][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-14 23:25:08,406][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-14 23:25:19,459][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-14 23:25:20,505][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 10:08:09,851][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 10:08:12,611][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-07-15/00-45-53/.hydra/config.yaml b/outputs/2026-07-15/00-45-53/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ce242abe530a0a180db7f1628c3b60cffcd95b66 --- /dev/null +++ b/outputs/2026-07-15/00-45-53/.hydra/config.yaml @@ -0,0 +1,238 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: false + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: true + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +name: train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k/ +seed: 42 +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-15/00-45-53/.hydra/hydra.yaml b/outputs/2026-07-15/00-45-53/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..251139a95ce9a5c69b806236943fde38f8bb4f0a --- /dev/null +++ b/outputs/2026-07-15/00-45-53/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k/ + - +seed=42 + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - experiment.num_nodes=1 + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=false + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=true + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.validation.inference_mode=false + - experiment.validation.val_every_n_step=2000 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k/,+seed=42,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_age.enabled=false,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=true,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.use_memory_attention=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.inference_mode=false,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-15/00-45-53 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-15/00-45-53/.hydra/overrides.yaml b/outputs/2026-07-15/00-45-53/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4d16bb2f69fa3890671e56a11c4fb41f680f0282 --- /dev/null +++ b/outputs/2026-07-15/00-45-53/.hydra/overrides.yaml @@ -0,0 +1,69 @@ +- +name=train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k/ +- +seed=42 +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- experiment.num_nodes=1 +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=false +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=true +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.validation.inference_mode=false +- experiment.validation.val_every_n_step=2000 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-07-15/00-45-53/main.log b/outputs/2026-07-15/00-45-53/main.log new file mode 100644 index 0000000000000000000000000000000000000000..fa0b384860a48ca6d35597f477fd70cf344368c1 --- /dev/null +++ b/outputs/2026-07-15/00-45-53/main.log @@ -0,0 +1,186 @@ +[2026-07-15 00:46:27,937][lightning.fabric.utilities.seed][INFO] - [rank: 6] Seed set to 42 +[2026-07-15 00:46:28,076][lightning.fabric.utilities.seed][INFO] - [rank: 4] Seed set to 42 +[2026-07-15 00:46:28,143][lightning.fabric.utilities.seed][INFO] - [rank: 3] Seed set to 42 +[2026-07-15 00:46:28,470][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-15 00:46:53,787][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-07-15 00:46:53,788][pytorch_lightning.utilities.rank_zero][INFO] - Missing checkpoint keys: spatial_rotary_emb.freqs, temporal_rotary_emb.freqs, pose_embedder.model.0.weight, pose_embedder.model.0.bias, pose_embedder.model.2.weight, pose_embedder.model.2.bias, memory_age_embed.weight, blocks.0.r_mlp.fc1.weight, ... (+301 more) +[2026-07-15 00:46:53,872][pytorch_lightning.utilities.rank_zero][INFO] - Zeroed frame-memory gate slices for base checkpoint load: blocks.0.r_adaLN_modulation.1.weight, blocks.0.r_adaLN_modulation.1.bias, blocks.1.r_adaLN_modulation.1.weight, blocks.1.r_adaLN_modulation.1.bias, blocks.2.r_adaLN_modulation.1.weight, blocks.2.r_adaLN_modulation.1.bias, blocks.3.r_adaLN_modulation.1.weight, blocks.3.r_adaLN_modulation.1.bias, ... (+24 more) +[2026-07-15 00:46:53,873][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: 302 tensors. +[2026-07-15 00:46:53,941][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-07-15 00:46:53,973][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: 228 tensors. +[2026-07-15 00:46:54,037][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-15 00:46:54,171][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-15 00:46:54,171][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-15 00:46:54,171][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-15 00:46:54,171][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-15 00:46:54,172][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-15 00:47:29,266][lightning.fabric.utilities.seed][INFO] - [rank: 6] Seed set to 42 +[2026-07-15 00:47:29,272][lightning.fabric.utilities.seed][INFO] - [rank: 4] Seed set to 42 +[2026-07-15 00:47:29,274][lightning.fabric.utilities.seed][INFO] - [rank: 3] Seed set to 42 +[2026-07-15 00:47:29,274][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-15 00:47:29,293][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-07-15 00:47:29,294][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-07-15 00:47:29,298][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-07-15 00:47:30,328][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-07-15 00:47:35,081][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-07-15 00:47:40,421][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 00:47:40,421][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 00:47:40,421][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 00:47:40,421][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 00:47:42,360][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +302 M Trainable params +839 M Non-trainable params +1.1 B Total params +4,569.428 Total estimated model params size (MB) +[2026-07-15 00:47:43,039][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 00:47:43,039][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 00:47:43,039][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 00:47:43,039][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 01:26:54,374][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,478][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,478][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,478][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,478][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,478][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,479][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,479][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,479][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,479][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,479][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,479][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,479][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,479][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,479][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,480][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,480][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,480][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 01:27:06,480][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-15 04:04:06,814][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=10000` reached. +[2026-07-15 04:04:20,620][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-15 04:04:20,622][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-15 04:04:20,622][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-15 04:04:20,622][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-15 04:04:20,622][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-15 04:04:22,056][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-15 04:04:22,057][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-15 04:04:22,057][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-15 04:04:22,057][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-15 04:04:22,057][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-15 04:04:46,824][lightning.fabric.utilities.seed][INFO] - [rank: 3] Seed set to 42 +[2026-07-15 04:04:46,900][lightning.fabric.utilities.seed][INFO] - [rank: 6] Seed set to 42 +[2026-07-15 04:04:46,987][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-15 04:04:47,299][lightning.fabric.utilities.seed][INFO] - [rank: 4] Seed set to 42 +[2026-07-15 04:04:47,414][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/callbacks/model_checkpoint.py:639: Checkpoint directory /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k/checkpoints exists and is not empty. + +[2026-07-15 04:04:47,414][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k/checkpoints/curriculum_01_no_updown_memory_only_step10000.ckpt +[2026-07-15 04:04:47,422][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-15 04:04:47,422][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-15 04:04:47,426][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-15 04:04:47,430][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-15 04:05:00,328][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 04:05:00,328][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 04:05:00,328][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 04:05:00,331][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 04:05:01,053][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +910 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,569.428 Total estimated model params size (MB) +[2026-07-15 04:05:01,390][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k/checkpoints/curriculum_01_no_updown_memory_only_step10000.ckpt +[2026-07-15 04:05:01,396][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 04:05:01,396][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 04:05:01,396][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 04:05:01,396][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 04:05:01,844][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/loops/training_epoch_loop.py:154: You're resuming from a checkpoint that ended before the epoch ended. This can cause unreliable results if further training is done. Consider using an end-of-epoch checkpoint + +[2026-07-15 15:04:22,147][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=40000` reached. +[2026-07-15 15:04:43,999][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-15 15:04:46,496][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-15 15:04:46,497][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-15 15:04:46,497][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-15 15:04:46,497][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-15 15:04:46,497][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-15 15:06:32,098][lightning.fabric.utilities.seed][INFO] - [rank: 6] Seed set to 42 +[2026-07-15 15:06:32,225][lightning.fabric.utilities.seed][INFO] - [rank: 4] Seed set to 42 +[2026-07-15 15:06:32,232][lightning.fabric.utilities.seed][INFO] - [rank: 3] Seed set to 42 +[2026-07-15 15:06:32,599][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-15 15:06:32,608][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k/checkpoints/curriculum_02_no_updown_step40000.ckpt +[2026-07-15 15:06:53,799][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 15:06:54,155][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 15:06:54,196][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 15:06:54,291][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 15:06:54,998][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +910 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,569.428 Total estimated model params size (MB) +[2026-07-15 15:06:55,966][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k/checkpoints/curriculum_02_no_updown_step40000.ckpt +[2026-07-15 15:06:55,990][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 15:06:55,990][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 15:06:55,990][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 15:06:55,991][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 00:46:05,086][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-07-18 00:46:05,086][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-07-18 00:46:05,087][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-18 00:46:05,087][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-18 00:46:05,086][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-07-18 00:46:05,087][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-18 00:46:05,086][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-07-18 00:46:05,088][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-18 00:46:05,090][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-07-18 00:46:05,089][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-07-18 00:46:05,090][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-18 00:46:05,090][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-07-18 00:46:05,090][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-18 00:46:05,090][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-18 00:46:05,090][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-07-18 00:46:05,090][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-07-15/00-45-54/.hydra/config.yaml b/outputs/2026-07-15/00-45-54/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ce242abe530a0a180db7f1628c3b60cffcd95b66 --- /dev/null +++ b/outputs/2026-07-15/00-45-54/.hydra/config.yaml @@ -0,0 +1,238 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: false + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: true + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +name: train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k/ +seed: 42 +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-15/00-45-54/.hydra/hydra.yaml b/outputs/2026-07-15/00-45-54/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a565637b259997ca68e6bb7820ddb3d205698ecf --- /dev/null +++ b/outputs/2026-07-15/00-45-54/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k/ + - +seed=42 + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - experiment.num_nodes=1 + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=false + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=true + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.validation.inference_mode=false + - experiment.validation.val_every_n_step=2000 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k/,+seed=42,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_age.enabled=false,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=true,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.use_memory_attention=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.inference_mode=false,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-15/00-45-54 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-15/00-45-54/.hydra/overrides.yaml b/outputs/2026-07-15/00-45-54/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4d16bb2f69fa3890671e56a11c4fb41f680f0282 --- /dev/null +++ b/outputs/2026-07-15/00-45-54/.hydra/overrides.yaml @@ -0,0 +1,69 @@ +- +name=train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_age_off_8h200_bs16_global128_seed42_200k/ +- +seed=42 +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- experiment.num_nodes=1 +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=false +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=true +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.validation.inference_mode=false +- experiment.validation.val_every_n_step=2000 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-07-15/00-45-54/main.log b/outputs/2026-07-15/00-45-54/main.log new file mode 100644 index 0000000000000000000000000000000000000000..92c97390815c6e7ed21932fc052576d95ecdc786 --- /dev/null +++ b/outputs/2026-07-15/00-45-54/main.log @@ -0,0 +1,76 @@ +[2026-07-15 00:46:27,938][lightning.fabric.utilities.seed][INFO] - [rank: 5] Seed set to 42 +[2026-07-15 00:46:28,240][lightning.fabric.utilities.seed][INFO] - [rank: 1] Seed set to 42 +[2026-07-15 00:46:28,416][lightning.fabric.utilities.seed][INFO] - [rank: 2] Seed set to 42 +[2026-07-15 00:46:28,804][lightning.fabric.utilities.seed][INFO] - [rank: 7] Seed set to 42 +[2026-07-15 00:47:29,264][lightning.fabric.utilities.seed][INFO] - [rank: 5] Seed set to 42 +[2026-07-15 00:47:29,265][lightning.fabric.utilities.seed][INFO] - [rank: 1] Seed set to 42 +[2026-07-15 00:47:29,266][lightning.fabric.utilities.seed][INFO] - [rank: 2] Seed set to 42 +[2026-07-15 00:47:29,465][lightning.fabric.utilities.seed][INFO] - [rank: 7] Seed set to 42 +[2026-07-15 00:47:29,475][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-07-15 00:47:30,332][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-07-15 00:47:31,081][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-07-15 00:47:31,327][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-07-15 00:47:40,421][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 00:47:40,421][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 00:47:40,421][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 00:47:40,421][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 00:47:43,039][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 00:47:43,039][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 00:47:43,039][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 00:47:43,039][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 04:04:20,620][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-15 04:04:20,622][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-15 04:04:20,622][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-15 04:04:20,622][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-15 04:04:46,836][lightning.fabric.utilities.seed][INFO] - [rank: 2] Seed set to 42 +[2026-07-15 04:04:46,870][lightning.fabric.utilities.seed][INFO] - [rank: 5] Seed set to 42 +[2026-07-15 04:04:46,905][lightning.fabric.utilities.seed][INFO] - [rank: 1] Seed set to 42 +[2026-07-15 04:04:46,970][lightning.fabric.utilities.seed][INFO] - [rank: 7] Seed set to 42 +[2026-07-15 04:04:47,424][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-15 04:04:47,428][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-15 04:04:47,432][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-15 04:04:47,434][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-15 04:05:00,320][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 04:05:00,322][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 04:05:00,325][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 04:05:00,329][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 04:05:01,396][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 04:05:01,396][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 04:05:01,396][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 04:05:01,396][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 15:06:32,111][lightning.fabric.utilities.seed][INFO] - [rank: 5] Seed set to 42 +[2026-07-15 15:06:32,153][lightning.fabric.utilities.seed][INFO] - [rank: 1] Seed set to 42 +[2026-07-15 15:06:32,379][lightning.fabric.utilities.seed][INFO] - [rank: 2] Seed set to 42 +[2026-07-15 15:06:32,566][lightning.fabric.utilities.seed][INFO] - [rank: 7] Seed set to 42 +[2026-07-15 15:06:54,020][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 15:06:54,165][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 15:06:54,211][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 15:06:54,236][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-15 15:06:55,990][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 15:06:55,990][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 15:06:55,990][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-15 15:06:55,990][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 00:46:05,086][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-07-18 00:46:05,086][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-07-18 00:46:05,086][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-07-18 00:46:05,087][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-18 00:46:05,087][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-18 00:46:05,087][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-18 00:46:05,086][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-07-18 00:46:05,087][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-18 00:46:05,090][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-07-18 00:46:05,090][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-07-18 00:46:05,090][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-18 00:46:05,090][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-18 00:46:05,090][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-07-18 00:46:05,090][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-18 00:46:05,104][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-07-18 00:46:05,104][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-07-17/18-47-36/.hydra/config.yaml b/outputs/2026-07-17/18-47-36/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..84c6a1f71a3a62f073b8f7e0bac7385cb564e117 --- /dev/null +++ b/outputs/2026-07-17/18-47-36/.hydra/config.yaml @@ -0,0 +1,238 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +name: train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/ +seed: 42 +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-17/18-47-36/.hydra/hydra.yaml b/outputs/2026-07-17/18-47-36/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a74a939d034f465a320cf31142d5d0c44b20d11d --- /dev/null +++ b/outputs/2026-07-17/18-47-36/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/ + - +seed=42 + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - experiment.num_nodes=1 + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.validation.inference_mode=false + - experiment.validation.val_every_n_step=2000 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/,+seed=42,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.use_memory_attention=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.inference_mode=false,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-17/18-47-36 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-17/18-47-36/.hydra/overrides.yaml b/outputs/2026-07-17/18-47-36/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5cad5dbdebf3980c008fde4c3141f3091d8d312c --- /dev/null +++ b/outputs/2026-07-17/18-47-36/.hydra/overrides.yaml @@ -0,0 +1,69 @@ +- +name=train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/ +- +seed=42 +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- experiment.num_nodes=1 +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.validation.inference_mode=false +- experiment.validation.val_every_n_step=2000 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-07-17/18-47-36/main.log b/outputs/2026-07-17/18-47-36/main.log new file mode 100644 index 0000000000000000000000000000000000000000..1e93952aa6fd0ebec728e23240169115b9b76050 --- /dev/null +++ b/outputs/2026-07-17/18-47-36/main.log @@ -0,0 +1,45 @@ +[2026-07-17 18:48:04,501][lightning.fabric.utilities.seed][INFO] - [rank: 1] Seed set to 42 +[2026-07-17 18:48:04,724][lightning.fabric.utilities.seed][INFO] - [rank: 7] Seed set to 42 +[2026-07-17 18:48:05,197][lightning.fabric.utilities.seed][INFO] - [rank: 4] Seed set to 42 +[2026-07-17 18:48:44,269][lightning.fabric.utilities.seed][INFO] - [rank: 4] Seed set to 42 +[2026-07-17 18:48:44,271][lightning.fabric.utilities.seed][INFO] - [rank: 1] Seed set to 42 +[2026-07-17 18:48:44,299][lightning.fabric.utilities.seed][INFO] - [rank: 7] Seed set to 42 +[2026-07-17 18:48:44,311][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-07-17 18:48:44,749][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-07-17 18:48:45,625][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-07-17 18:49:08,784][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 18:49:08,784][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 18:49:08,784][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 18:49:11,454][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 18:49:11,454][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 18:49:11,454][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 22:03:14,599][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-17 22:03:14,600][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-17 22:03:14,600][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-17 22:03:43,027][lightning.fabric.utilities.seed][INFO] - [rank: 1] Seed set to 42 +[2026-07-17 22:03:43,136][lightning.fabric.utilities.seed][INFO] - [rank: 7] Seed set to 42 +[2026-07-17 22:03:43,246][lightning.fabric.utilities.seed][INFO] - [rank: 4] Seed set to 42 +[2026-07-17 22:03:43,373][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-17 22:03:43,377][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-17 22:03:43,384][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-17 22:03:46,430][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 22:03:46,436][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 22:03:46,782][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 22:03:47,808][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 22:03:47,808][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 22:03:47,808][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 08:55:31,666][lightning.fabric.utilities.seed][INFO] - [rank: 1] Seed set to 42 +[2026-07-18 08:55:31,765][lightning.fabric.utilities.seed][INFO] - [rank: 4] Seed set to 42 +[2026-07-18 08:55:32,203][lightning.fabric.utilities.seed][INFO] - [rank: 7] Seed set to 42 +[2026-07-18 08:55:49,607][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 08:55:49,672][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 08:55:49,778][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 08:55:51,468][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 08:55:51,468][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 08:55:51,469][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. diff --git a/outputs/2026-07-17/18-47-37/.hydra/config.yaml b/outputs/2026-07-17/18-47-37/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..84c6a1f71a3a62f073b8f7e0bac7385cb564e117 --- /dev/null +++ b/outputs/2026-07-17/18-47-37/.hydra/config.yaml @@ -0,0 +1,238 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +name: train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/ +seed: 42 +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-17/18-47-37/.hydra/hydra.yaml b/outputs/2026-07-17/18-47-37/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..35839a28e30c70030bb89e21761c7bb563ada069 --- /dev/null +++ b/outputs/2026-07-17/18-47-37/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/ + - +seed=42 + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - experiment.num_nodes=1 + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.validation.inference_mode=false + - experiment.validation.val_every_n_step=2000 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/,+seed=42,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.use_memory_attention=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.inference_mode=false,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-17/18-47-37 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-17/18-47-37/.hydra/overrides.yaml b/outputs/2026-07-17/18-47-37/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5cad5dbdebf3980c008fde4c3141f3091d8d312c --- /dev/null +++ b/outputs/2026-07-17/18-47-37/.hydra/overrides.yaml @@ -0,0 +1,69 @@ +- +name=train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/ +- +seed=42 +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- experiment.num_nodes=1 +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.validation.inference_mode=false +- experiment.validation.val_every_n_step=2000 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-07-17/18-47-37/main.log b/outputs/2026-07-17/18-47-37/main.log new file mode 100644 index 0000000000000000000000000000000000000000..699289a144171934efc33ac2e2f3da94ba935239 --- /dev/null +++ b/outputs/2026-07-17/18-47-37/main.log @@ -0,0 +1,186 @@ +[2026-07-17 18:48:04,461][lightning.fabric.utilities.seed][INFO] - [rank: 2] Seed set to 42 +[2026-07-17 18:48:04,580][lightning.fabric.utilities.seed][INFO] - [rank: 3] Seed set to 42 +[2026-07-17 18:48:04,746][lightning.fabric.utilities.seed][INFO] - [rank: 6] Seed set to 42 +[2026-07-17 18:48:04,874][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-17 18:48:04,914][lightning.fabric.utilities.seed][INFO] - [rank: 5] Seed set to 42 +[2026-07-17 18:48:29,146][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-07-17 18:48:29,147][pytorch_lightning.utilities.rank_zero][INFO] - Missing checkpoint keys: spatial_rotary_emb.freqs, temporal_rotary_emb.freqs, pose_embedder.model.0.weight, pose_embedder.model.0.bias, pose_embedder.model.2.weight, pose_embedder.model.2.bias, memory_age_embed.weight, blocks.0.r_mlp.fc1.weight, ... (+301 more) +[2026-07-17 18:48:29,234][pytorch_lightning.utilities.rank_zero][INFO] - Zeroed frame-memory gate slices for base checkpoint load: blocks.0.r_adaLN_modulation.1.weight, blocks.0.r_adaLN_modulation.1.bias, blocks.1.r_adaLN_modulation.1.weight, blocks.1.r_adaLN_modulation.1.bias, blocks.2.r_adaLN_modulation.1.weight, blocks.2.r_adaLN_modulation.1.bias, blocks.3.r_adaLN_modulation.1.weight, blocks.3.r_adaLN_modulation.1.bias, ... (+24 more) +[2026-07-17 18:48:29,235][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: 302 tensors. +[2026-07-17 18:48:29,327][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-07-17 18:48:29,364][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: 228 tensors. +[2026-07-17 18:48:29,420][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-17 18:48:29,561][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-17 18:48:29,561][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-17 18:48:29,561][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-17 18:48:29,561][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-17 18:48:29,562][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-17 18:48:44,269][lightning.fabric.utilities.seed][INFO] - [rank: 2] Seed set to 42 +[2026-07-17 18:48:44,278][lightning.fabric.utilities.seed][INFO] - [rank: 5] Seed set to 42 +[2026-07-17 18:48:44,294][lightning.fabric.utilities.seed][INFO] - [rank: 3] Seed set to 42 +[2026-07-17 18:48:44,296][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-07-17 18:48:44,305][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-07-17 18:48:44,898][lightning.fabric.utilities.seed][INFO] - [rank: 6] Seed set to 42 +[2026-07-17 18:48:44,913][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-07-17 18:48:44,921][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-07-17 18:48:57,501][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-17 18:48:57,522][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-07-17 18:49:03,203][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-07-17 18:49:08,784][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 18:49:08,784][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 18:49:08,784][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 18:49:08,784][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 18:49:08,784][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 18:49:10,463][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +302 M Trainable params +839 M Non-trainable params +1.1 B Total params +4,569.428 Total estimated model params size (MB) +[2026-07-17 18:49:11,454][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 18:49:11,454][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 18:49:11,454][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 18:49:11,454][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 18:49:11,454][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 19:27:46,772][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,444][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,444][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,444][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,445][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,445][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,445][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,445][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,445][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,445][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,445][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,445][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,445][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,446][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,446][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,446][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,446][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,446][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 19:27:59,446][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 22:03:02,002][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=10000` reached. +[2026-07-17 22:03:14,599][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-17 22:03:14,600][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-17 22:03:14,600][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-17 22:03:14,600][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-17 22:03:14,600][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-17 22:03:14,600][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-17 22:03:14,911][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-17 22:03:14,912][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-17 22:03:14,912][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-17 22:03:14,912][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-17 22:03:14,912][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-17 22:03:43,082][lightning.fabric.utilities.seed][INFO] - [rank: 2] Seed set to 42 +[2026-07-17 22:03:43,092][lightning.fabric.utilities.seed][INFO] - [rank: 3] Seed set to 42 +[2026-07-17 22:03:43,115][lightning.fabric.utilities.seed][INFO] - [rank: 5] Seed set to 42 +[2026-07-17 22:03:43,119][lightning.fabric.utilities.seed][INFO] - [rank: 6] Seed set to 42 +[2026-07-17 22:03:43,208][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-17 22:03:43,363][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/callbacks/model_checkpoint.py:639: Checkpoint directory /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints exists and is not empty. + +[2026-07-17 22:03:43,363][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/curriculum_01_no_updown_memory_only_step10000.ckpt +[2026-07-17 22:03:43,375][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-17 22:03:43,379][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-17 22:03:43,381][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-17 22:03:43,384][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-17 22:03:43,385][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-17 22:03:46,431][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 22:03:46,432][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 22:03:46,435][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 22:03:46,438][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 22:03:46,593][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 22:03:47,465][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +910 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,569.428 Total estimated model params size (MB) +[2026-07-17 22:03:47,799][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/curriculum_01_no_updown_memory_only_step10000.ckpt +[2026-07-17 22:03:47,808][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 22:03:47,808][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 22:03:47,808][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 22:03:47,808][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 22:03:47,808][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 22:03:48,308][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/loops/training_epoch_loop.py:154: You're resuming from a checkpoint that ended before the epoch ended. This can cause unreliable results if further training is done. Consider using an end-of-epoch checkpoint + +[2026-07-18 08:53:54,423][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=40000` reached. +[2026-07-18 08:54:16,527][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-18 08:54:17,349][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-18 08:54:17,349][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-18 08:54:17,349][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-18 08:54:17,349][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-18 08:54:17,349][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-18 08:55:31,648][lightning.fabric.utilities.seed][INFO] - [rank: 3] Seed set to 42 +[2026-07-18 08:55:31,676][lightning.fabric.utilities.seed][INFO] - [rank: 6] Seed set to 42 +[2026-07-18 08:55:31,774][lightning.fabric.utilities.seed][INFO] - [rank: 5] Seed set to 42 +[2026-07-18 08:55:31,779][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-18 08:55:32,070][lightning.fabric.utilities.seed][INFO] - [rank: 2] Seed set to 42 +[2026-07-18 08:55:32,212][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/curriculum_02_no_updown_step40000.ckpt +[2026-07-18 08:55:49,006][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 08:55:49,466][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 08:55:49,482][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 08:55:49,628][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 08:55:49,717][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 08:55:50,470][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +910 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,569.428 Total estimated model params size (MB) +[2026-07-18 08:55:51,448][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/curriculum_02_no_updown_step40000.ckpt +[2026-07-18 08:55:51,468][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 08:55:51,468][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 08:55:51,468][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 08:55:51,468][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 08:55:51,468][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-20 18:20:12,133][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=200000` reached. diff --git a/outputs/2026-07-17/20-20-55/.hydra/config.yaml b/outputs/2026-07-17/20-20-55/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..1a7fb5307bee9d6a98cf01c653319c9750f51439 --- /dev/null +++ b/outputs/2026-07-17/20-20-55/.hydra/config.yaml @@ -0,0 +1,238 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: false + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: true + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +name: train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k/ +seed: 42 +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-17/20-20-55/.hydra/hydra.yaml b/outputs/2026-07-17/20-20-55/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c26123117e9f3ce494420437601abbcb8b5d8cc8 --- /dev/null +++ b/outputs/2026-07-17/20-20-55/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k/ + - +seed=42 + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - experiment.num_nodes=1 + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=true + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=false + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.validation.inference_mode=false + - experiment.validation.val_every_n_step=2000 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k/,+seed=42,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=false,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=true,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.use_memory_attention=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.inference_mode=false,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-17/20-20-55 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-17/20-20-55/.hydra/overrides.yaml b/outputs/2026-07-17/20-20-55/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..132373229e24f12a173c6dd4448ca44267af9b52 --- /dev/null +++ b/outputs/2026-07-17/20-20-55/.hydra/overrides.yaml @@ -0,0 +1,69 @@ +- +name=train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k/ +- +seed=42 +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- experiment.num_nodes=1 +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=true +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=false +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.validation.inference_mode=false +- experiment.validation.val_every_n_step=2000 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-07-17/20-20-55/main.log b/outputs/2026-07-17/20-20-55/main.log new file mode 100644 index 0000000000000000000000000000000000000000..ccd4191fafc7f8cb540fdc902ec2125c09cf13c3 --- /dev/null +++ b/outputs/2026-07-17/20-20-55/main.log @@ -0,0 +1,212 @@ +[2026-07-17 20:21:33,079][lightning.fabric.utilities.seed][INFO] - [rank: 4] Seed set to 42 +[2026-07-17 20:21:33,079][lightning.fabric.utilities.seed][INFO] - [rank: 2] Seed set to 42 +[2026-07-17 20:21:33,111][lightning.fabric.utilities.seed][INFO] - [rank: 6] Seed set to 42 +[2026-07-17 20:21:33,332][lightning.fabric.utilities.seed][INFO] - [rank: 5] Seed set to 42 +[2026-07-17 20:21:33,386][lightning.fabric.utilities.seed][INFO] - [rank: 1] Seed set to 42 +[2026-07-17 20:21:33,594][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-17 20:21:59,990][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-07-17 20:21:59,991][pytorch_lightning.utilities.rank_zero][INFO] - Missing checkpoint keys: spatial_rotary_emb.freqs, temporal_rotary_emb.freqs, pose_embedder.model.0.weight, pose_embedder.model.0.bias, pose_embedder.model.2.weight, pose_embedder.model.2.bias, memory_age_embed.weight, blocks.0.r_mlp.fc1.weight, ... (+301 more) +[2026-07-17 20:22:00,079][pytorch_lightning.utilities.rank_zero][INFO] - Zeroed frame-memory gate slices for base checkpoint load: blocks.0.r_adaLN_modulation.1.weight, blocks.0.r_adaLN_modulation.1.bias, blocks.1.r_adaLN_modulation.1.weight, blocks.1.r_adaLN_modulation.1.bias, blocks.2.r_adaLN_modulation.1.weight, blocks.2.r_adaLN_modulation.1.bias, blocks.3.r_adaLN_modulation.1.weight, blocks.3.r_adaLN_modulation.1.bias, ... (+24 more) +[2026-07-17 20:22:00,080][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: 302 tensors. +[2026-07-17 20:22:00,125][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-07-17 20:22:00,155][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: 228 tensors. +[2026-07-17 20:22:00,195][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-17 20:22:00,341][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-17 20:22:00,341][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-17 20:22:00,341][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-17 20:22:00,341][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-17 20:22:00,342][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-17 20:22:36,732][lightning.fabric.utilities.seed][INFO] - [rank: 5] Seed set to 42 +[2026-07-17 20:22:36,733][lightning.fabric.utilities.seed][INFO] - [rank: 6] Seed set to 42 +[2026-07-17 20:22:36,733][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-17 20:22:36,736][lightning.fabric.utilities.seed][INFO] - [rank: 4] Seed set to 42 +[2026-07-17 20:22:36,742][lightning.fabric.utilities.seed][INFO] - [rank: 2] Seed set to 42 +[2026-07-17 20:22:36,762][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-07-17 20:22:36,763][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-07-17 20:22:37,126][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-07-17 20:22:37,575][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-07-17 20:22:37,581][lightning.fabric.utilities.seed][INFO] - [rank: 1] Seed set to 42 +[2026-07-17 20:22:37,595][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-07-17 20:22:37,619][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-07-17 20:22:43,016][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-07-17 20:22:47,961][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 20:22:47,961][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 20:22:47,961][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 20:22:47,961][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 20:22:47,961][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 20:22:47,961][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 20:22:49,570][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +302 M Trainable params +839 M Non-trainable params +1.1 B Total params +4,569.428 Total estimated model params size (MB) +[2026-07-17 20:22:50,925][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 20:22:50,925][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 20:22:50,925][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 20:22:50,925][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 20:22:50,925][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 20:22:50,926][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 21:02:06,335][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,113][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,113][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,113][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,113][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,114][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,114][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,114][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,114][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,114][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,114][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,114][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,114][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,114][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,115][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,115][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,115][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,115][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 21:02:12,115][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-17 23:39:10,050][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=10000` reached. +[2026-07-17 23:39:20,875][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-17 23:39:20,875][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-17 23:39:20,875][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-17 23:39:20,875][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-17 23:39:20,876][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-17 23:39:20,876][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-17 23:39:20,876][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-17 23:39:21,142][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-17 23:39:21,142][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-17 23:39:21,143][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-17 23:39:21,143][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-17 23:39:21,143][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-17 23:39:49,503][lightning.fabric.utilities.seed][INFO] - [rank: 6] Seed set to 42 +[2026-07-17 23:39:49,523][lightning.fabric.utilities.seed][INFO] - [rank: 2] Seed set to 42 +[2026-07-17 23:39:49,555][lightning.fabric.utilities.seed][INFO] - [rank: 5] Seed set to 42 +[2026-07-17 23:39:49,571][lightning.fabric.utilities.seed][INFO] - [rank: 4] Seed set to 42 +[2026-07-17 23:39:49,661][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-17 23:39:50,265][lightning.fabric.utilities.seed][INFO] - [rank: 1] Seed set to 42 +[2026-07-17 23:39:50,381][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/callbacks/model_checkpoint.py:639: Checkpoint directory /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k/checkpoints exists and is not empty. + +[2026-07-17 23:39:50,382][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k/checkpoints/curriculum_01_no_updown_memory_only_step10000.ckpt +[2026-07-17 23:39:50,386][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-17 23:39:50,396][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-17 23:39:50,399][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-17 23:39:50,401][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-17 23:39:50,403][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-17 23:39:50,403][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-17 23:40:02,153][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 23:40:02,155][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 23:40:02,158][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 23:40:02,158][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 23:40:02,159][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 23:40:02,164][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 23:40:02,848][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +910 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,569.428 Total estimated model params size (MB) +[2026-07-17 23:40:03,205][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k/checkpoints/curriculum_01_no_updown_memory_only_step10000.ckpt +[2026-07-17 23:40:03,208][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 23:40:03,208][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 23:40:03,208][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 23:40:03,208][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 23:40:03,208][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 23:40:03,208][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 23:40:03,628][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/loops/training_epoch_loop.py:154: You're resuming from a checkpoint that ended before the epoch ended. This can cause unreliable results if further training is done. Consider using an end-of-epoch checkpoint + +[2026-07-18 10:36:55,971][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=40000` reached. +[2026-07-18 10:37:16,851][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-18 10:37:19,186][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-18 10:37:19,186][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-18 10:37:19,186][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-18 10:37:19,186][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-18 10:37:19,186][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-18 10:39:05,856][lightning.fabric.utilities.seed][INFO] - [rank: 6] Seed set to 42 +[2026-07-18 10:39:05,873][lightning.fabric.utilities.seed][INFO] - [rank: 1] Seed set to 42 +[2026-07-18 10:39:05,915][lightning.fabric.utilities.seed][INFO] - [rank: 2] Seed set to 42 +[2026-07-18 10:39:05,967][lightning.fabric.utilities.seed][INFO] - [rank: 4] Seed set to 42 +[2026-07-18 10:39:06,036][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-18 10:39:06,184][lightning.fabric.utilities.seed][INFO] - [rank: 5] Seed set to 42 +[2026-07-18 10:39:06,281][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k/checkpoints/curriculum_02_no_updown_step40000.ckpt +[2026-07-18 10:39:22,483][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 10:39:22,751][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 10:39:22,788][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 10:39:22,865][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 10:39:22,965][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 10:39:23,036][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 10:39:23,734][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +910 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,569.428 Total estimated model params size (MB) +[2026-07-18 10:39:24,697][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k/checkpoints/curriculum_02_no_updown_step40000.ckpt +[2026-07-18 10:39:24,790][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 10:39:24,790][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 10:39:24,790][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 10:39:24,791][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 10:39:24,791][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 10:39:24,791][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-20 20:20:56,441][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-07-20 20:20:56,442][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-20 20:20:56,441][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-07-20 20:20:56,441][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-07-20 20:20:56,442][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-20 20:20:56,441][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-07-20 20:20:56,443][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-20 20:20:56,443][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-20 20:20:56,441][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-07-20 20:20:56,441][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-07-20 20:20:56,444][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-20 20:20:56,444][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-07-17/20-20-56/.hydra/config.yaml b/outputs/2026-07-17/20-20-56/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..1a7fb5307bee9d6a98cf01c653319c9750f51439 --- /dev/null +++ b/outputs/2026-07-17/20-20-56/.hydra/config.yaml @@ -0,0 +1,238 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: false + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: true + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +name: train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k/ +seed: 42 +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-17/20-20-56/.hydra/hydra.yaml b/outputs/2026-07-17/20-20-56/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f50827871b721fbdcc1f74c6abfa6883a7719f3d --- /dev/null +++ b/outputs/2026-07-17/20-20-56/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k/ + - +seed=42 + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - experiment.num_nodes=1 + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=true + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=false + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.validation.inference_mode=false + - experiment.validation.val_every_n_step=2000 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k/,+seed=42,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=false,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=true,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.use_memory_attention=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.inference_mode=false,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-17/20-20-56 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-17/20-20-56/.hydra/overrides.yaml b/outputs/2026-07-17/20-20-56/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..132373229e24f12a173c6dd4448ca44267af9b52 --- /dev/null +++ b/outputs/2026-07-17/20-20-56/.hydra/overrides.yaml @@ -0,0 +1,69 @@ +- +name=train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_noise_off_8h200_bs16_global128_seed42_200k/ +- +seed=42 +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- experiment.num_nodes=1 +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=true +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=false +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.validation.inference_mode=false +- experiment.validation.val_every_n_step=2000 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-07-17/20-20-56/main.log b/outputs/2026-07-17/20-20-56/main.log new file mode 100644 index 0000000000000000000000000000000000000000..61e02592d9245acbe19c4bbfeb414af4a5b265a2 --- /dev/null +++ b/outputs/2026-07-17/20-20-56/main.log @@ -0,0 +1,34 @@ +[2026-07-17 20:21:33,313][lightning.fabric.utilities.seed][INFO] - [rank: 7] Seed set to 42 +[2026-07-17 20:21:33,466][lightning.fabric.utilities.seed][INFO] - [rank: 3] Seed set to 42 +[2026-07-17 20:22:36,732][lightning.fabric.utilities.seed][INFO] - [rank: 3] Seed set to 42 +[2026-07-17 20:22:36,841][lightning.fabric.utilities.seed][INFO] - [rank: 7] Seed set to 42 +[2026-07-17 20:22:36,853][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-07-17 20:22:38,075][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-07-17 20:22:47,961][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 20:22:47,961][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 20:22:50,925][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 20:22:50,925][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 23:39:20,875][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-17 23:39:20,876][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-17 23:39:49,532][lightning.fabric.utilities.seed][INFO] - [rank: 3] Seed set to 42 +[2026-07-17 23:39:49,567][lightning.fabric.utilities.seed][INFO] - [rank: 7] Seed set to 42 +[2026-07-17 23:39:50,392][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-17 23:39:50,394][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-17 23:40:02,159][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 23:40:02,159][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-17 23:40:03,208][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-17 23:40:03,208][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 10:39:05,943][lightning.fabric.utilities.seed][INFO] - [rank: 7] Seed set to 42 +[2026-07-18 10:39:06,033][lightning.fabric.utilities.seed][INFO] - [rank: 3] Seed set to 42 +[2026-07-18 10:39:22,638][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 10:39:23,010][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 10:39:24,790][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 10:39:24,791][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-20 20:20:56,441][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-07-20 20:20:56,441][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-07-20 20:20:56,443][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-20 20:20:56,443][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-07-18/00-46-53/.hydra/config.yaml b/outputs/2026-07-18/00-46-53/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75ad480d630dbab8223b46ed26a1a096ea62d1cf --- /dev/null +++ b/outputs/2026-07-18/00-46-53/.hydra/config.yaml @@ -0,0 +1,238 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: true + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +name: train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k/ +seed: 42 +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-18/00-46-53/.hydra/hydra.yaml b/outputs/2026-07-18/00-46-53/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3dff69be154cee587303713e7464dd6778bbe4d7 --- /dev/null +++ b/outputs/2026-07-18/00-46-53/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k/ + - +seed=42 + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - experiment.num_nodes=1 + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=true + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.validation.inference_mode=false + - experiment.validation.val_every_n_step=2000 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k/,+seed=42,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=true,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.use_memory_attention=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.inference_mode=false,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-18/00-46-53 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-18/00-46-53/.hydra/overrides.yaml b/outputs/2026-07-18/00-46-53/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..56018cf8195f9be151815a3c8e6f6e4e8fa0508c --- /dev/null +++ b/outputs/2026-07-18/00-46-53/.hydra/overrides.yaml @@ -0,0 +1,69 @@ +- +name=train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k/ +- +seed=42 +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- experiment.num_nodes=1 +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=true +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.validation.inference_mode=false +- experiment.validation.val_every_n_step=2000 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-07-18/00-46-53/main.log b/outputs/2026-07-18/00-46-53/main.log new file mode 100644 index 0000000000000000000000000000000000000000..cdaea20c6c196f3648ef8cc3630c1647451ed278 --- /dev/null +++ b/outputs/2026-07-18/00-46-53/main.log @@ -0,0 +1,195 @@ +[2026-07-18 00:47:16,254][lightning.fabric.utilities.seed][INFO] - [rank: 1] Seed set to 42 +[2026-07-18 00:47:16,295][lightning.fabric.utilities.seed][INFO] - [rank: 7] Seed set to 42 +[2026-07-18 00:47:16,457][lightning.fabric.utilities.seed][INFO] - [rank: 2] Seed set to 42 +[2026-07-18 00:47:16,491][lightning.fabric.utilities.seed][INFO] - [rank: 5] Seed set to 42 +[2026-07-18 00:47:16,972][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-18 00:47:39,241][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-07-18 00:47:39,242][pytorch_lightning.utilities.rank_zero][INFO] - Missing checkpoint keys: spatial_rotary_emb.freqs, temporal_rotary_emb.freqs, pose_embedder.model.0.weight, pose_embedder.model.0.bias, pose_embedder.model.2.weight, pose_embedder.model.2.bias, memory_age_embed.weight, blocks.0.r_mlp.fc1.weight, ... (+301 more) +[2026-07-18 00:47:39,325][pytorch_lightning.utilities.rank_zero][INFO] - Zeroed frame-memory gate slices for base checkpoint load: blocks.0.r_adaLN_modulation.1.weight, blocks.0.r_adaLN_modulation.1.bias, blocks.1.r_adaLN_modulation.1.weight, blocks.1.r_adaLN_modulation.1.bias, blocks.2.r_adaLN_modulation.1.weight, blocks.2.r_adaLN_modulation.1.bias, blocks.3.r_adaLN_modulation.1.weight, blocks.3.r_adaLN_modulation.1.bias, ... (+24 more) +[2026-07-18 00:47:39,325][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors: 302 tensors. +[2026-07-18 00:47:39,396][pytorch_lightning.utilities.rank_zero][INFO] - Checkpoint prefix match: strip '', add ''. +[2026-07-18 00:47:39,430][pytorch_lightning.utilities.rank_zero][INFO] - Model weights loaded from /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors: 228 tensors. +[2026-07-18 00:47:39,480][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-18 00:47:39,609][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-18 00:47:39,610][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-18 00:47:39,610][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-18 00:47:39,610][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-18 00:47:39,610][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-18 00:47:53,255][lightning.fabric.utilities.seed][INFO] - [rank: 1] Seed set to 42 +[2026-07-18 00:47:53,255][lightning.fabric.utilities.seed][INFO] - [rank: 7] Seed set to 42 +[2026-07-18 00:47:53,255][lightning.fabric.utilities.seed][INFO] - [rank: 2] Seed set to 42 +[2026-07-18 00:47:53,257][lightning.fabric.utilities.seed][INFO] - [rank: 5] Seed set to 42 +[2026-07-18 00:47:53,277][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8 +[2026-07-18 00:47:54,186][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8 +[2026-07-18 00:47:54,393][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8 +[2026-07-18 00:47:55,206][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8 +[2026-07-18 00:48:04,621][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-18 00:48:04,639][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8 +[2026-07-18 00:48:10,725][pytorch_lightning.utilities.rank_zero][INFO] - ---------------------------------------------------------------------------------------------------- +distributed_backend=nccl +All distributed processes registered. Starting with 8 processes +---------------------------------------------------------------------------------------------------- + +[2026-07-18 00:48:16,106][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 00:48:16,106][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 00:48:16,106][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 00:48:16,107][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 00:48:16,107][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 00:48:18,583][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +302 M Trainable params +839 M Non-trainable params +1.1 B Total params +4,569.428 Total estimated model params size (MB) +[2026-07-18 00:48:19,589][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 00:48:19,589][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 00:48:19,589][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 00:48:19,590][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 00:48:19,590][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 01:27:32,956][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('validation/latent_mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,171][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('mse', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,172][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('psnr', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,172][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('lpips', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,172][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,172][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,172][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0000_0100', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,172][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,173][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,173][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0100_0200', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,173][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,173][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,173][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0200_0300', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,173][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,173][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,173][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0300_0400', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,173][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/mse_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,173][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/psnr_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 01:27:38,174][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/logger_connector/result.py:433: It is recommended to use `self.log('segment/lpips_0400_0500', ..., sync_dist=True)` when logging on epoch level in distributed setting to accumulate the metric across devices. + +[2026-07-18 04:04:28,192][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=10000` reached. +[2026-07-18 04:04:40,544][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-18 04:04:40,544][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-18 04:04:40,545][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-18 04:04:40,545][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-18 04:04:40,545][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-18 04:04:40,545][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-18 04:04:43,168][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-18 04:04:43,168][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-18 04:04:43,168][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-18 04:04:43,168][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-18 04:04:43,169][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-18 04:05:08,312][lightning.fabric.utilities.seed][INFO] - [rank: 7] Seed set to 42 +[2026-07-18 04:05:08,315][lightning.fabric.utilities.seed][INFO] - [rank: 5] Seed set to 42 +[2026-07-18 04:05:08,329][lightning.fabric.utilities.seed][INFO] - [rank: 2] Seed set to 42 +[2026-07-18 04:05:08,494][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-18 04:05:08,786][lightning.fabric.utilities.seed][INFO] - [rank: 1] Seed set to 42 +[2026-07-18 04:05:08,900][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/callbacks/model_checkpoint.py:639: Checkpoint directory /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k/checkpoints exists and is not empty. + +[2026-07-18 04:05:08,901][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k/checkpoints/curriculum_01_no_updown_memory_only_step10000.ckpt +[2026-07-18 04:05:08,909][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-18 04:05:08,918][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-18 04:05:08,926][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-18 04:05:08,927][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-18 04:05:08,928][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-18 04:05:20,355][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 04:05:20,357][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 04:05:20,357][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 04:05:20,362][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 04:05:20,367][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 04:05:21,129][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +910 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,569.428 Total estimated model params size (MB) +[2026-07-18 04:05:21,457][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k/checkpoints/curriculum_01_no_updown_memory_only_step10000.ckpt +[2026-07-18 04:05:21,468][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 04:05:21,468][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 04:05:21,468][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 04:05:21,468][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 04:05:21,468][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 04:05:21,945][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/loops/training_epoch_loop.py:154: You're resuming from a checkpoint that ended before the epoch ended. This can cause unreliable results if further training is done. Consider using an end-of-epoch checkpoint + +[2026-07-18 15:03:52,032][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer.fit` stopped: `max_steps=40000` reached. +[2026-07-18 15:04:12,865][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-18 15:04:13,105][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-18 15:04:13,105][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-18 15:04:13,105][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-18 15:04:13,106][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-18 15:04:13,106][pytorch_lightning.utilities.rank_zero][INFO] - `Trainer(limit_val_batches=1)` was configured so 1 batch will be used. +[2026-07-18 15:05:23,418][lightning.fabric.utilities.seed][INFO] - [rank: 1] Seed set to 42 +[2026-07-18 15:05:23,524][lightning.fabric.utilities.seed][INFO] - [rank: 7] Seed set to 42 +[2026-07-18 15:05:23,875][lightning.fabric.utilities.seed][INFO] - [rank: 2] Seed set to 42 +[2026-07-18 15:05:24,240][lightning.fabric.utilities.seed][INFO] - [rank: 5] Seed set to 42 +[2026-07-18 15:05:24,345][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-18 15:05:24,353][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k/checkpoints/curriculum_02_no_updown_step40000.ckpt +[2026-07-18 15:05:43,351][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 2 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 15:05:43,411][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 15:05:43,537][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 1 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 15:05:43,564][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 5 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 15:05:43,605][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 7 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 15:05:44,310][lightning.pytorch.callbacks.model_summary][INFO] - + | Name | Type | Params +--------------------------------------------------------------------------------- +0 | diffusion_model | Diffusion | 910 M +1 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M +2 | vae | AutoencoderKL | 229 M +--------------------------------------------------------------------------------- +910 M Trainable params +231 M Non-trainable params +1.1 B Total params +4,569.428 Total estimated model params size (MB) +[2026-07-18 15:05:45,287][pytorch_lightning.utilities.rank_zero][INFO] - Restored all states from the checkpoint at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k/checkpoints/curriculum_02_no_updown_step40000.ckpt +[2026-07-18 15:05:45,310][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 15:05:45,310][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 15:05:45,310][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 15:05:45,310][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 15:05:45,311][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-21 00:47:00,364][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15 +[2026-07-21 00:47:00,366][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-21 00:47:00,364][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15 +[2026-07-21 00:47:00,364][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15 +[2026-07-21 00:47:00,367][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-21 00:47:00,367][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-21 00:47:00,364][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15 +[2026-07-21 00:47:00,368][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-21 00:47:00,364][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15 +[2026-07-21 00:47:00,368][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/2026-07-18/00-46-54/.hydra/config.yaml b/outputs/2026-07-18/00-46-54/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75ad480d630dbab8223b46ed26a1a096ea62d1cf --- /dev/null +++ b/outputs/2026-07-18/00-46-54/.hydra/config.yaml @@ -0,0 +1,238 @@ +experiment: + debug: ${debug} + tasks: + - training + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 16 + max_epochs: -1 + max_steps: 200000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2000 + every_n_epochs: null + train_time_interval: null + save_last_k: 5 + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 40000 + dataset: + wo_updown: true + - name: with_updown + until_step: 200000 + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2000 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 500 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: true + full_dit_start_step: 10000 + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: true + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: true + save_local: false + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +auto_resume: true +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +name: train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k +output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k/ +seed: 42 +customized_load: true +seperate_load: true +diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors diff --git a/outputs/2026-07-18/00-46-54/.hydra/hydra.yaml b/outputs/2026-07-18/00-46-54/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..1cb88eb5b24513b380b6921379532858e109e2c5 --- /dev/null +++ b/outputs/2026-07-18/00-46-54/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S} + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.mode=RUN + task: + - +name=train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k + - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k/ + - +seed=42 + - auto_resume=true + - wandb.mode=offline + - experiment.tasks=[training] + - experiment.num_nodes=1 + - algorithm=dememwm_base + - +customized_load=true + - +seperate_load=true + - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors + - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=true + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.warmup_steps=500 + - algorithm.trainability.train_full_dit=true + - algorithm.trainability.full_dit_start_step=10000 + - algorithm.trainability.lr.memory_modules=8.0e-5 + - algorithm.trainability.lr.base_dit=2.0e-5 + - algorithm.log_video=true + - algorithm.save_local=false + - algorithm.diffusion.sampling_timesteps=20 + - experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} + - experiment.training.batch_size=16 + - experiment.training.optim.accumulate_grad_batches=1 + - experiment.validation.batch_size=1 + - experiment.validation.limit_batch=1 + - experiment.validation.inference_mode=false + - experiment.validation.val_every_n_step=2000 + - experiment.training.checkpointing.every_n_train_steps=2000 + - experiment.training.checkpointing.save_last_k=5 + - experiment.training.max_steps=200000 + job: + name: main + chdir: null + override_dirname: +customized_load=true,+diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors,+name=train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k,+output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k/,+seed=42,+seperate_load=true,+vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=true,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=true,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=false,algorithm.trainability.full_dit_start_step=10000,algorithm.trainability.lr.base_dit=2.0e-5,algorithm.trainability.lr.memory_modules=8.0e-5,algorithm.trainability.train_full_dit=true,algorithm.use_memory_attention=true,algorithm.warmup_steps=500,algorithm=dememwm_base,auto_resume=true,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[training],experiment.training.batch_size=16,experiment.training.checkpointing.every_n_train_steps=2000,experiment.training.checkpointing.save_last_k=5,experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]},experiment.training.max_steps=200000,experiment.training.optim.accumulate_grad_batches=1,experiment.validation.batch_size=1,experiment.validation.inference_mode=false,experiment.validation.limit_batch=1,experiment.validation.val_every_n_step=2000,wandb.mode=offline + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-07-18/00-46-54 + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/2026-07-18/00-46-54/.hydra/overrides.yaml b/outputs/2026-07-18/00-46-54/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..56018cf8195f9be151815a3c8e6f6e4e8fa0508c --- /dev/null +++ b/outputs/2026-07-18/00-46-54/.hydra/overrides.yaml @@ -0,0 +1,69 @@ +- +name=train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k +- +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_ref_8h200_bs16_global128_seed42_200k/ +- +seed=42 +- auto_resume=true +- wandb.mode=offline +- experiment.tasks=[training] +- experiment.num_nodes=1 +- algorithm=dememwm_base +- +customized_load=true +- +seperate_load=true +- +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors +- +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=true +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.warmup_steps=500 +- algorithm.trainability.train_full_dit=true +- algorithm.trainability.full_dit_start_step=10000 +- algorithm.trainability.lr.memory_modules=8.0e-5 +- algorithm.trainability.lr.base_dit=2.0e-5 +- algorithm.log_video=true +- algorithm.save_local=false +- algorithm.diffusion.sampling_timesteps=20 +- experiment.training.curriculum={enabled:true,stages:[{name:no_updown,until_step:40000,dataset:{wo_updown:true}},{name:with_updown,until_step:200000,dataset:{wo_updown:false}}]} +- experiment.training.batch_size=16 +- experiment.training.optim.accumulate_grad_batches=1 +- experiment.validation.batch_size=1 +- experiment.validation.limit_batch=1 +- experiment.validation.inference_mode=false +- experiment.validation.val_every_n_step=2000 +- experiment.training.checkpointing.every_n_train_steps=2000 +- experiment.training.checkpointing.save_last_k=5 +- experiment.training.max_steps=200000 diff --git a/outputs/2026-07-18/00-46-54/main.log b/outputs/2026-07-18/00-46-54/main.log new file mode 100644 index 0000000000000000000000000000000000000000..768b94577291d7d01d0c3e2addb2cd282534cd08 --- /dev/null +++ b/outputs/2026-07-18/00-46-54/main.log @@ -0,0 +1,51 @@ +[2026-07-18 00:47:16,194][lightning.fabric.utilities.seed][INFO] - [rank: 3] Seed set to 42 +[2026-07-18 00:47:16,493][lightning.fabric.utilities.seed][INFO] - [rank: 6] Seed set to 42 +[2026-07-18 00:47:16,806][lightning.fabric.utilities.seed][INFO] - [rank: 4] Seed set to 42 +[2026-07-18 00:47:53,255][lightning.fabric.utilities.seed][INFO] - [rank: 4] Seed set to 42 +[2026-07-18 00:47:53,261][lightning.fabric.utilities.seed][INFO] - [rank: 6] Seed set to 42 +[2026-07-18 00:47:53,270][lightning.fabric.utilities.seed][INFO] - [rank: 3] Seed set to 42 +[2026-07-18 00:47:53,270][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8 +[2026-07-18 00:47:53,280][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8 +[2026-07-18 00:47:53,850][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8 +[2026-07-18 00:48:16,106][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 00:48:16,106][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 00:48:16,106][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 00:48:19,590][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 00:48:19,590][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 00:48:19,590][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 04:04:40,545][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-18 04:04:40,545][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-18 04:04:40,544][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/plugins/precision/amp.py:54: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + +[2026-07-18 04:05:08,320][lightning.fabric.utilities.seed][INFO] - [rank: 4] Seed set to 42 +[2026-07-18 04:05:08,335][lightning.fabric.utilities.seed][INFO] - [rank: 6] Seed set to 42 +[2026-07-18 04:05:08,350][lightning.fabric.utilities.seed][INFO] - [rank: 3] Seed set to 42 +[2026-07-18 04:05:08,911][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-18 04:05:08,913][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-18 04:05:08,927][py.warnings][WARNING] - /proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/fabric/utilities/cloud_io.py:56: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + +[2026-07-18 04:05:20,358][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 04:05:20,360][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 04:05:20,367][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 04:05:21,468][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 04:05:21,468][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 04:05:21,468][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 15:05:23,489][lightning.fabric.utilities.seed][INFO] - [rank: 6] Seed set to 42 +[2026-07-18 15:05:23,502][lightning.fabric.utilities.seed][INFO] - [rank: 4] Seed set to 42 +[2026-07-18 15:05:23,751][lightning.fabric.utilities.seed][INFO] - [rank: 3] Seed set to 42 +[2026-07-18 15:05:43,451][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 6 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 15:05:43,520][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 3 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 15:05:43,572][lightning.pytorch.accelerators.cuda][INFO] - LOCAL_RANK: 4 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7] +[2026-07-18 15:05:45,310][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 15:05:45,310][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-18 15:05:45,310][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers. +[2026-07-21 00:47:00,364][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15 +[2026-07-21 00:47:00,364][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15 +[2026-07-21 00:47:00,366][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-21 00:47:00,366][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 +[2026-07-21 00:47:00,364][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15 +[2026-07-21 00:47:00,368][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15 diff --git a/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/checkpoint_sha256.csv b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/checkpoint_sha256.csv new file mode 100644 index 0000000000000000000000000000000000000000..3fe398b65aa346cbb5d4e527eedee75ce919868a --- /dev/null +++ b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/checkpoint_sha256.csv @@ -0,0 +1,2 @@ +path,size_bytes,sha256 +/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/checkpoints/epoch3_step200000.ckpt,11853122333,9db3ec83065c8ba004305b89e6db2bbcf4c5e96c940b1f169de858cc73e31c30 diff --git a/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/evaluation.patch b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/evaluation.patch new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.csv b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..3cfd08bc49f4159d2bd5cace6409ac92956d78e1 --- /dev/null +++ b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.csv @@ -0,0 +1,51 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.sha256 b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..67b0dc2042229c0768d324282215bdf64bebe28f --- /dev/null +++ b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.sha256 @@ -0,0 +1 @@ +f9536911e434610cce23306ff99575db53cd7fc29e3df841da4b6631bf418634 diff --git a/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/git_provenance.json b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/git_provenance.json new file mode 100644 index 0000000000000000000000000000000000000000..fabe691fc08806f957f2445d4a52b2f38d7f07d8 --- /dev/null +++ b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/git_provenance.json @@ -0,0 +1,5 @@ +{ + "branch": "dememwm-unified-sourcefree-age-balanced", + "head": "732555281e86c4fdbb71c98a000d648c3edc3109", + "status": "?? results/\n?? train_dememwm_full_berzelius_gate_off_h200.sh\n?? train_dememwm_separate_mem_gatefix_4layers512_valid500_berzelius_h200.sh" +} diff --git a/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/protocol.json b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/protocol.json new file mode 100644 index 0000000000000000000000000000000000000000..67066ccb30c3755ec1c0ab8f1663cbd1c4dabc31 --- /dev/null +++ b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/protocol.json @@ -0,0 +1,122 @@ +{ + "model": "dememwm_step200k", + "master_seed": 42, + "sample_seed_formula": "master_seed * 1000000 + global_test_index", + "test_clips": 300, + "source_offset": 100, + "context_range": [ + 100, + 700 + ], + "target_range": [ + 700, + 1200 + ], + "context_frames": 600, + "generated_frames": 500, + "local_window": 8, + "ddim_steps": 20, + "ddim_eta": 0.0, + "per_device_batch_size": 1, + "worker_ranges": [ + "0:10", + "10:20", + "20:30", + "30:40", + "40:50", + "50:60", + "60:70", + "70:80", + "80:90", + "90:100", + "100:110", + "110:120", + "120:129", + "129:138", + "138:147", + "147:156", + "156:165", + "165:174", + "174:183", + "183:192", + "192:201", + "201:210", + "210:219", + "219:228", + "228:237", + "237:246", + "246:255", + "255:264", + "264:273", + "273:282", + "282:291", + "291:300" + ], + "distribution": "32 disjoint standalone batch-1 workers; four per GPU; no DistributedSampler", + "generated_frame_bins": [ + [ + 0, + 100 + ], + [ + 100, + 200 + ], + [ + 200, + 300 + ], + [ + 300, + 400 + ], + [ + 400, + 500 + ] + ], + "reconstruction_metrics": { + "names": [ + "mse", + "psnr", + "lpips" + ], + "population_per_bin": { + "clips": 300, + "frames": 30000 + }, + "population_overall": { + "clips": 300, + "frames": 150000 + }, + "lpips_input_range": [ + 0.0, + 1.0 + ], + "lpips_normalize": true + }, + "frame_fid": { + "implementation": "TorchMetrics FrechetInceptionDistance", + "feature": 2048, + "normalize": false, + "input": "uint8 RGB [0,255]", + "official_compatible_bins": { + "clips": 50, + "frames_per_bin": 5000 + }, + "supplementary_pooled": { + "clips": 50, + "frames": 25000 + }, + "full_300_bins": { + "clips": 300, + "frames_per_bin": 30000 + }, + "full_300_pooled": { + "clips": 300, + "frames": 150000 + } + }, + "manifest_sha256": "ef4e4de9d3975d8a7e724bb609e57f20186cd1f7285457c00fb85c1aca973b99", + "fid50_manifest_sha256": "f9536911e434610cce23306ff99575db53cd7fc29e3df841da4b6631bf418634" +} diff --git a/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.csv b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..dae70b4071c2d2a834268c2a26f061d308f11f2c --- /dev/null +++ b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.csv @@ -0,0 +1,301 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +50,50,212,test/val_savanna/1/000039.mp4,test/val_savanna/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +51,51,4,test/val_desert/0/000001.mp4,test/val_desert/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +52,52,7,test/val_desert/0/000002.mp4,test/val_desert/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +53,53,310,test/val_iceplains/0/000026.mp4,test/val_iceplains/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +54,54,284,test/val_iceplains/0/000015.mp4,test/val_iceplains/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +55,55,109,test/val_desert/3/000041.mp4,test/val_desert/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +56,56,45,test/val_desert/0/000018.mp4,test/val_desert/0/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +57,57,93,test/val_desert/0/000035.mp4,test/val_desert/0/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +58,58,400,test/val_plains/3/000011.mp4,test/val_plains/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +59,59,300,test/val_iceplains/2/000021.mp4,test/val_iceplains/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +60,60,226,test/val_savanna/2/000043.mp4,test/val_savanna/2/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +61,61,417,test/val_plains/2/000018.mp4,test/val_plains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +62,62,120,test/val_desert/3/000046.mp4,test/val_desert/3/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +63,63,106,test/val_desert/3/000040.mp4,test/val_desert/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +64,64,345,test/val_iceplains/1/000039.mp4,test/val_iceplains/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +65,65,169,test/val_savanna/2/000016.mp4,test/val_savanna/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +66,66,312,test/val_iceplains/1/000027.mp4,test/val_iceplains/1/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +67,67,368,test/val_iceplains/0/000048.mp4,test/val_iceplains/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +68,68,449,test/val_plains/1/000034.mp4,test/val_plains/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +69,69,11,test/val_desert/1/000004.mp4,test/val_desert/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +70,70,6,test/val_desert/2/000001.mp4,test/val_desert/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +71,71,257,test/val_iceplains/0/000005.mp4,test/val_iceplains/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +72,72,262,test/val_iceplains/3/000006.mp4,test/val_iceplains/3/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +73,73,164,test/val_savanna/0/000015.mp4,test/val_savanna/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +74,74,337,test/val_iceplains/2/000036.mp4,test/val_iceplains/2/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +75,75,95,test/val_desert/0/000036.mp4,test/val_desert/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +76,76,277,test/val_iceplains/0/000013.mp4,test/val_iceplains/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +77,77,26,test/val_desert/1/000010.mp4,test/val_desert/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +78,78,77,test/val_desert/0/000029.mp4,test/val_desert/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +79,79,294,test/val_iceplains/1/000019.mp4,test/val_iceplains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +80,80,34,test/val_desert/1/000014.mp4,test/val_desert/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +81,81,188,test/val_savanna/1/000029.mp4,test/val_savanna/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +82,82,170,test/val_savanna/0/000017.mp4,test/val_savanna/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +83,83,415,test/val_plains/2/000017.mp4,test/val_plains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +84,84,145,test/val_savanna/0/000006.mp4,test/val_savanna/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +85,85,133,test/val_savanna/1/000002.mp4,test/val_savanna/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +86,86,347,test/val_iceplains/2/000039.mp4,test/val_iceplains/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +87,87,209,test/val_savanna/0/000038.mp4,test/val_savanna/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +88,88,404,test/val_plains/1/000013.mp4,test/val_plains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +89,89,211,test/val_savanna/2/000038.mp4,test/val_savanna/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +90,90,464,test/val_plains/2/000044.mp4,test/val_plains/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +91,91,162,test/val_savanna/1/000014.mp4,test/val_savanna/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +92,92,419,test/val_plains/2/000019.mp4,test/val_plains/2/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +93,93,127,test/val_desert/3/000049.mp4,test/val_desert/3/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +94,94,370,test/val_iceplains/2/000048.mp4,test/val_iceplains/2/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +95,95,293,test/val_iceplains/2/000018.mp4,test/val_iceplains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +96,96,251,test/val_iceplains/1/000003.mp4,test/val_iceplains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +97,97,441,test/val_plains/1/000031.mp4,test/val_plains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +98,98,69,test/val_desert/2/000026.mp4,test/val_desert/2/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +99,99,286,test/val_iceplains/2/000015.mp4,test/val_iceplains/2/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +100,100,438,test/val_plains/0/000029.mp4,test/val_plains/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +101,101,376,test/val_plains/0/000002.mp4,test/val_plains/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +102,102,163,test/val_savanna/2/000014.mp4,test/val_savanna/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +103,103,31,test/val_desert/1/000013.mp4,test/val_desert/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +104,104,241,test/val_savanna/0/000049.mp4,test/val_savanna/0/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +105,105,458,test/val_plains/0/000040.mp4,test/val_plains/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +106,106,371,test/val_iceplains/1/000049.mp4,test/val_iceplains/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +107,107,230,test/val_savanna/3/000045.mp4,test/val_savanna/3/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +108,108,372,test/val_iceplains/2/000049.mp4,test/val_iceplains/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +109,109,223,test/val_savanna/1/000042.mp4,test/val_savanna/1/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +110,110,68,test/val_desert/0/000026.mp4,test/val_desert/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +111,111,461,test/val_plains/2/000042.mp4,test/val_plains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +112,112,151,test/val_savanna/1/000009.mp4,test/val_savanna/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +113,113,72,test/val_desert/3/000027.mp4,test/val_desert/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +114,114,427,test/val_plains/0/000024.mp4,test/val_plains/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +115,115,317,test/val_iceplains/3/000029.mp4,test/val_iceplains/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +116,116,249,test/val_iceplains/2/000001.mp4,test/val_iceplains/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +117,117,364,test/val_iceplains/2/000046.mp4,test/val_iceplains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +118,118,96,test/val_desert/2/000037.mp4,test/val_desert/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +119,119,190,test/val_savanna/3/000029.mp4,test/val_savanna/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +120,120,227,test/val_savanna/2/000044.mp4,test/val_savanna/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +121,121,244,test/val_iceplains/0/000000.mp4,test/val_iceplains/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +122,122,38,test/val_desert/3/000015.mp4,test/val_desert/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +123,123,291,test/val_iceplains/2/000017.mp4,test/val_iceplains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +124,124,85,test/val_desert/2/000032.mp4,test/val_desert/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +125,125,358,test/val_iceplains/0/000044.mp4,test/val_iceplains/0/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +126,126,50,test/val_desert/1/000020.mp4,test/val_desert/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +127,127,207,test/val_savanna/0/000036.mp4,test/val_savanna/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +128,128,330,test/val_iceplains/2/000034.mp4,test/val_iceplains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +129,129,132,test/val_savanna/0/000001.mp4,test/val_savanna/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +130,130,423,test/val_plains/3/000021.mp4,test/val_plains/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +131,131,66,test/val_desert/2/000025.mp4,test/val_desert/2/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +132,132,153,test/val_savanna/3/000009.mp4,test/val_savanna/3/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +133,133,382,test/val_plains/0/000004.mp4,test/val_plains/0/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +134,134,143,test/val_savanna/2/000005.mp4,test/val_savanna/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +135,135,348,test/val_iceplains/1/000040.mp4,test/val_iceplains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +136,136,384,test/val_plains/2/000005.mp4,test/val_plains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +137,137,30,test/val_desert/0/000012.mp4,test/val_desert/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +138,138,203,test/val_savanna/2/000034.mp4,test/val_savanna/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +139,139,179,test/val_savanna/0/000022.mp4,test/val_savanna/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +140,140,253,test/val_iceplains/2/000003.mp4,test/val_iceplains/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +141,141,289,test/val_iceplains/2/000016.mp4,test/val_iceplains/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +142,142,422,test/val_plains/2/000020.mp4,test/val_plains/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +143,143,20,test/val_desert/0/000008.mp4,test/val_desert/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +144,144,130,test/val_savanna/0/000000.mp4,test/val_savanna/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +145,145,296,test/val_iceplains/1/000020.mp4,test/val_iceplains/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +146,146,278,test/val_iceplains/3/000013.mp4,test/val_iceplains/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +147,147,269,test/val_iceplains/0/000009.mp4,test/val_iceplains/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +148,148,314,test/val_iceplains/0/000028.mp4,test/val_iceplains/0/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +149,149,281,test/val_iceplains/0/000014.mp4,test/val_iceplains/0/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +150,150,149,test/val_savanna/0/000008.mp4,test/val_savanna/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +151,151,184,test/val_savanna/0/000024.mp4,test/val_savanna/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +152,152,407,test/val_plains/1/000014.mp4,test/val_plains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +153,153,304,test/val_iceplains/2/000023.mp4,test/val_iceplains/2/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +154,154,208,test/val_savanna/3/000037.mp4,test/val_savanna/3/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +155,155,182,test/val_savanna/3/000023.mp4,test/val_savanna/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +156,156,394,test/val_plains/1/000009.mp4,test/val_plains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +157,157,14,test/val_desert/0/000006.mp4,test/val_desert/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +158,158,55,test/val_desert/3/000021.mp4,test/val_desert/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +159,159,75,test/val_desert/2/000028.mp4,test/val_desert/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +160,160,319,test/val_iceplains/1/000031.mp4,test/val_iceplains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +161,161,39,test/val_desert/1/000016.mp4,test/val_desert/1/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +162,162,418,test/val_plains/1/000019.mp4,test/val_plains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +163,163,389,test/val_plains/3/000007.mp4,test/val_plains/3/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +164,164,409,test/val_plains/2/000014.mp4,test/val_plains/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +165,165,344,test/val_iceplains/2/000038.mp4,test/val_iceplains/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +166,166,396,test/val_plains/2/000009.mp4,test/val_plains/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +167,167,67,test/val_desert/1/000026.mp4,test/val_desert/1/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +168,168,99,test/val_desert/3/000038.mp4,test/val_desert/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +169,169,288,test/val_iceplains/3/000016.mp4,test/val_iceplains/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +170,170,97,test/val_desert/1/000038.mp4,test/val_desert/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +171,171,178,test/val_savanna/2/000021.mp4,test/val_savanna/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +172,172,264,test/val_iceplains/1/000007.mp4,test/val_iceplains/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +173,173,266,test/val_iceplains/0/000008.mp4,test/val_iceplains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +174,174,303,test/val_iceplains/1/000023.mp4,test/val_iceplains/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +175,175,442,test/val_plains/0/000031.mp4,test/val_plains/0/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +176,176,233,test/val_savanna/1/000047.mp4,test/val_savanna/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +177,177,240,test/val_savanna/1/000049.mp4,test/val_savanna/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +178,178,42,test/val_desert/0/000017.mp4,test/val_desert/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +179,179,369,test/val_iceplains/3/000048.mp4,test/val_iceplains/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +180,180,398,test/val_plains/3/000010.mp4,test/val_plains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +181,181,435,test/val_plains/2/000027.mp4,test/val_plains/2/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +182,182,242,test/val_savanna/2/000049.mp4,test/val_savanna/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +183,183,361,test/val_iceplains/2/000045.mp4,test/val_iceplains/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +184,184,470,test/val_plains/2/000047.mp4,test/val_plains/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +185,185,356,test/val_iceplains/2/000042.mp4,test/val_iceplains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +186,186,468,test/val_plains/2/000046.mp4,test/val_plains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +187,187,113,test/val_desert/0/000043.mp4,test/val_desert/0/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +188,188,19,test/val_desert/1/000008.mp4,test/val_desert/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +189,189,199,test/val_savanna/0/000033.mp4,test/val_savanna/0/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +190,190,118,test/val_desert/2/000045.mp4,test/val_desert/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +191,191,433,test/val_plains/0/000027.mp4,test/val_plains/0/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +192,192,328,test/val_iceplains/0/000034.mp4,test/val_iceplains/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +193,193,213,test/val_savanna/0/000039.mp4,test/val_savanna/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +194,194,339,test/val_iceplains/0/000037.mp4,test/val_iceplains/0/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +195,195,466,test/val_plains/1/000046.mp4,test/val_plains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +196,196,355,test/val_iceplains/3/000042.mp4,test/val_iceplains/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +197,197,336,test/val_iceplains/3/000036.mp4,test/val_iceplains/3/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +198,198,420,test/val_plains/0/000020.mp4,test/val_plains/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +199,199,410,test/val_plains/1/000015.mp4,test/val_plains/1/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +200,200,27,test/val_desert/1/000011.mp4,test/val_desert/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +201,201,86,test/val_desert/1/000033.mp4,test/val_desert/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +202,202,210,test/val_savanna/3/000038.mp4,test/val_savanna/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +203,203,103,test/val_desert/3/000039.mp4,test/val_desert/3/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +204,204,245,test/val_iceplains/3/000000.mp4,test/val_iceplains/3/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +205,205,25,test/val_desert/2/000009.mp4,test/val_desert/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +206,206,340,test/val_iceplains/2/000037.mp4,test/val_iceplains/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +207,207,167,test/val_savanna/0/000016.mp4,test/val_savanna/0/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +208,208,424,test/val_plains/0/000022.mp4,test/val_plains/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +209,209,448,test/val_plains/2/000033.mp4,test/val_plains/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +210,210,386,test/val_plains/0/000006.mp4,test/val_plains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +211,211,392,test/val_plains/3/000008.mp4,test/val_plains/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +212,212,105,test/val_desert/1/000040.mp4,test/val_desert/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +213,213,267,test/val_iceplains/2/000008.mp4,test/val_iceplains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +214,214,90,test/val_desert/0/000034.mp4,test/val_desert/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +215,215,426,test/val_plains/0/000023.mp4,test/val_plains/0/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +216,216,168,test/val_savanna/3/000016.mp4,test/val_savanna/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +217,217,218,test/val_savanna/3/000040.mp4,test/val_savanna/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +218,218,445,test/val_plains/0/000032.mp4,test/val_plains/0/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +219,219,48,test/val_desert/3/000019.mp4,test/val_desert/3/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +220,220,104,test/val_desert/2/000039.mp4,test/val_desert/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +221,221,206,test/val_savanna/2/000035.mp4,test/val_savanna/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +222,222,228,test/val_savanna/1/000045.mp4,test/val_savanna/1/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +223,223,200,test/val_savanna/2/000033.mp4,test/val_savanna/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +224,224,62,test/val_desert/2/000024.mp4,test/val_desert/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +225,225,121,test/val_desert/1/000047.mp4,test/val_desert/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +226,226,147,test/val_savanna/0/000007.mp4,test/val_savanna/0/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +227,227,146,test/val_savanna/2/000006.mp4,test/val_savanna/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +228,228,225,test/val_savanna/1/000043.mp4,test/val_savanna/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +229,229,313,test/val_iceplains/3/000027.mp4,test/val_iceplains/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +230,230,354,test/val_iceplains/0/000042.mp4,test/val_iceplains/0/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +231,231,367,test/val_iceplains/1/000048.mp4,test/val_iceplains/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +232,232,297,test/val_iceplains/3/000020.mp4,test/val_iceplains/3/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +233,233,124,test/val_desert/1/000048.mp4,test/val_desert/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +234,234,171,test/val_savanna/2/000017.mp4,test/val_savanna/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +235,235,342,test/val_iceplains/0/000038.mp4,test/val_iceplains/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +236,236,60,test/val_desert/3/000023.mp4,test/val_desert/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +237,237,131,test/val_savanna/1/000001.mp4,test/val_savanna/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +238,238,365,test/val_iceplains/1/000047.mp4,test/val_iceplains/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +239,239,315,test/val_iceplains/2/000028.mp4,test/val_iceplains/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +240,240,220,test/val_savanna/1/000041.mp4,test/val_savanna/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +241,241,8,test/val_desert/1/000003.mp4,test/val_desert/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +242,242,261,test/val_iceplains/0/000006.mp4,test/val_iceplains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +243,243,17,test/val_desert/1/000007.mp4,test/val_desert/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +244,244,462,test/val_plains/1/000043.mp4,test/val_plains/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +245,245,454,test/val_plains/3/000038.mp4,test/val_plains/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +246,246,351,test/val_iceplains/1/000041.mp4,test/val_iceplains/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +247,247,383,test/val_plains/3/000005.mp4,test/val_plains/3/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +248,248,18,test/val_desert/2/000007.mp4,test/val_desert/2/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +249,249,457,test/val_plains/1/000040.mp4,test/val_plains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +250,250,326,test/val_iceplains/3/000033.mp4,test/val_iceplains/3/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +251,251,254,test/val_iceplains/1/000004.mp4,test/val_iceplains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +252,252,375,test/val_plains/1/000002.mp4,test/val_plains/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +253,253,141,test/val_savanna/0/000005.mp4,test/val_savanna/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +254,254,306,test/val_iceplains/2/000024.mp4,test/val_iceplains/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +255,255,10,test/val_desert/2/000003.mp4,test/val_desert/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +256,256,173,test/val_savanna/0/000020.mp4,test/val_savanna/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +257,257,320,test/val_iceplains/1/000032.mp4,test/val_iceplains/1/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +258,258,115,test/val_desert/1/000044.mp4,test/val_desert/1/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +259,259,341,test/val_iceplains/1/000038.mp4,test/val_iceplains/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +260,260,61,test/val_desert/3/000024.mp4,test/val_desert/3/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +261,261,231,test/val_savanna/1/000046.mp4,test/val_savanna/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +262,262,224,test/val_savanna/3/000042.mp4,test/val_savanna/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +263,263,292,test/val_iceplains/3/000018.mp4,test/val_iceplains/3/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +264,264,334,test/val_iceplains/1/000036.mp4,test/val_iceplains/1/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +265,265,159,test/val_savanna/0/000013.mp4,test/val_savanna/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +266,266,102,test/val_desert/0/000039.mp4,test/val_desert/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +267,267,362,test/val_iceplains/1/000046.mp4,test/val_iceplains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +268,268,322,test/val_iceplains/3/000032.mp4,test/val_iceplains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +269,269,381,test/val_plains/1/000004.mp4,test/val_plains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +270,270,84,test/val_desert/3/000032.mp4,test/val_desert/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +271,271,243,test/val_iceplains/1/000000.mp4,test/val_iceplains/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +272,272,65,test/val_desert/3/000025.mp4,test/val_desert/3/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +273,273,64,test/val_desert/0/000025.mp4,test/val_desert/0/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +274,274,272,test/val_iceplains/2/000010.mp4,test/val_iceplains/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +275,275,399,test/val_plains/0/000011.mp4,test/val_plains/0/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +276,276,248,test/val_iceplains/3/000001.mp4,test/val_iceplains/3/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +277,277,43,test/val_desert/3/000017.mp4,test/val_desert/3/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +278,278,393,test/val_plains/2/000008.mp4,test/val_plains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +279,279,123,test/val_desert/2/000047.mp4,test/val_desert/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +280,280,29,test/val_desert/2/000011.mp4,test/val_desert/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +281,281,122,test/val_desert/0/000047.mp4,test/val_desert/0/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +282,282,157,test/val_savanna/3/000011.mp4,test/val_savanna/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +283,283,333,test/val_iceplains/2/000035.mp4,test/val_iceplains/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +284,284,307,test/val_iceplains/1/000025.mp4,test/val_iceplains/1/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +285,285,9,test/val_desert/0/000003.mp4,test/val_desert/0/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +286,286,250,test/val_iceplains/2/000002.mp4,test/val_iceplains/2/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +287,287,165,test/val_savanna/3/000015.mp4,test/val_savanna/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +288,288,0,test/val_desert/1/000000.mp4,test/val_desert/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +289,289,271,test/val_iceplains/3/000010.mp4,test/val_iceplains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +290,290,276,test/val_iceplains/1/000013.mp4,test/val_iceplains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +291,291,82,test/val_desert/1/000031.mp4,test/val_desert/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +292,292,191,test/val_savanna/2/000029.mp4,test/val_savanna/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +293,293,374,test/val_plains/1/000001.mp4,test/val_plains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +294,294,446,test/val_plains/3/000032.mp4,test/val_plains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +295,295,378,test/val_plains/1/000003.mp4,test/val_plains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +296,296,152,test/val_savanna/0/000009.mp4,test/val_savanna/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +297,297,54,test/val_desert/0/000021.mp4,test/val_desert/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +298,298,259,test/val_iceplains/2/000005.mp4,test/val_iceplains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +299,299,91,test/val_desert/2/000034.mp4,test/val_desert/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.sha256 b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..065350069e5b0f46e33cd51ad63f60aa956822fd --- /dev/null +++ b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.sha256 @@ -0,0 +1 @@ +ef4e4de9d3975d8a7e724bb609e57f20186cd1f7285457c00fb85c1aca973b99 diff --git a/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/checkpoint_sha256.csv b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/checkpoint_sha256.csv new file mode 100644 index 0000000000000000000000000000000000000000..3fe398b65aa346cbb5d4e527eedee75ce919868a --- /dev/null +++ b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/checkpoint_sha256.csv @@ -0,0 +1,2 @@ +path,size_bytes,sha256 +/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_unified_dense_memory_ddp_selection_geometry_8h200_bs16_global128_200k/checkpoints/epoch3_step200000.ckpt,11853122333,9db3ec83065c8ba004305b89e6db2bbcf4c5e96c940b1f169de858cc73e31c30 diff --git a/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/evaluation.patch b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/evaluation.patch new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.csv b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..3cfd08bc49f4159d2bd5cace6409ac92956d78e1 --- /dev/null +++ b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.csv @@ -0,0 +1,51 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.sha256 b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..67b0dc2042229c0768d324282215bdf64bebe28f --- /dev/null +++ b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.sha256 @@ -0,0 +1 @@ +f9536911e434610cce23306ff99575db53cd7fc29e3df841da4b6631bf418634 diff --git a/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/git_provenance.json b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/git_provenance.json new file mode 100644 index 0000000000000000000000000000000000000000..fabe691fc08806f957f2445d4a52b2f38d7f07d8 --- /dev/null +++ b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/git_provenance.json @@ -0,0 +1,5 @@ +{ + "branch": "dememwm-unified-sourcefree-age-balanced", + "head": "732555281e86c4fdbb71c98a000d648c3edc3109", + "status": "?? results/\n?? train_dememwm_full_berzelius_gate_off_h200.sh\n?? train_dememwm_separate_mem_gatefix_4layers512_valid500_berzelius_h200.sh" +} diff --git a/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/protocol.json b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/protocol.json new file mode 100644 index 0000000000000000000000000000000000000000..4a5133bf8fd69c74b0f92207278dcc941568a9f2 --- /dev/null +++ b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/protocol.json @@ -0,0 +1,122 @@ +{ + "model": "dememwm_step200k", + "master_seed": 43, + "sample_seed_formula": "master_seed * 1000000 + global_test_index", + "test_clips": 300, + "source_offset": 100, + "context_range": [ + 100, + 700 + ], + "target_range": [ + 700, + 1200 + ], + "context_frames": 600, + "generated_frames": 500, + "local_window": 8, + "ddim_steps": 20, + "ddim_eta": 0.0, + "per_device_batch_size": 1, + "worker_ranges": [ + "0:10", + "10:20", + "20:30", + "30:40", + "40:50", + "50:60", + "60:70", + "70:80", + "80:90", + "90:100", + "100:110", + "110:120", + "120:129", + "129:138", + "138:147", + "147:156", + "156:165", + "165:174", + "174:183", + "183:192", + "192:201", + "201:210", + "210:219", + "219:228", + "228:237", + "237:246", + "246:255", + "255:264", + "264:273", + "273:282", + "282:291", + "291:300" + ], + "distribution": "32 disjoint standalone batch-1 workers; four per GPU; no DistributedSampler", + "generated_frame_bins": [ + [ + 0, + 100 + ], + [ + 100, + 200 + ], + [ + 200, + 300 + ], + [ + 300, + 400 + ], + [ + 400, + 500 + ] + ], + "reconstruction_metrics": { + "names": [ + "mse", + "psnr", + "lpips" + ], + "population_per_bin": { + "clips": 300, + "frames": 30000 + }, + "population_overall": { + "clips": 300, + "frames": 150000 + }, + "lpips_input_range": [ + 0.0, + 1.0 + ], + "lpips_normalize": true + }, + "frame_fid": { + "implementation": "TorchMetrics FrechetInceptionDistance", + "feature": 2048, + "normalize": false, + "input": "uint8 RGB [0,255]", + "official_compatible_bins": { + "clips": 50, + "frames_per_bin": 5000 + }, + "supplementary_pooled": { + "clips": 50, + "frames": 25000 + }, + "full_300_bins": { + "clips": 300, + "frames_per_bin": 30000 + }, + "full_300_pooled": { + "clips": 300, + "frames": 150000 + } + }, + "manifest_sha256": "ef4e4de9d3975d8a7e724bb609e57f20186cd1f7285457c00fb85c1aca973b99", + "fid50_manifest_sha256": "f9536911e434610cce23306ff99575db53cd7fc29e3df841da4b6631bf418634" +} diff --git a/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.csv b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..dae70b4071c2d2a834268c2a26f061d308f11f2c --- /dev/null +++ b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.csv @@ -0,0 +1,301 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +50,50,212,test/val_savanna/1/000039.mp4,test/val_savanna/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +51,51,4,test/val_desert/0/000001.mp4,test/val_desert/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +52,52,7,test/val_desert/0/000002.mp4,test/val_desert/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +53,53,310,test/val_iceplains/0/000026.mp4,test/val_iceplains/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +54,54,284,test/val_iceplains/0/000015.mp4,test/val_iceplains/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +55,55,109,test/val_desert/3/000041.mp4,test/val_desert/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +56,56,45,test/val_desert/0/000018.mp4,test/val_desert/0/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +57,57,93,test/val_desert/0/000035.mp4,test/val_desert/0/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +58,58,400,test/val_plains/3/000011.mp4,test/val_plains/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +59,59,300,test/val_iceplains/2/000021.mp4,test/val_iceplains/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +60,60,226,test/val_savanna/2/000043.mp4,test/val_savanna/2/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +61,61,417,test/val_plains/2/000018.mp4,test/val_plains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +62,62,120,test/val_desert/3/000046.mp4,test/val_desert/3/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +63,63,106,test/val_desert/3/000040.mp4,test/val_desert/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +64,64,345,test/val_iceplains/1/000039.mp4,test/val_iceplains/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +65,65,169,test/val_savanna/2/000016.mp4,test/val_savanna/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +66,66,312,test/val_iceplains/1/000027.mp4,test/val_iceplains/1/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +67,67,368,test/val_iceplains/0/000048.mp4,test/val_iceplains/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +68,68,449,test/val_plains/1/000034.mp4,test/val_plains/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +69,69,11,test/val_desert/1/000004.mp4,test/val_desert/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +70,70,6,test/val_desert/2/000001.mp4,test/val_desert/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +71,71,257,test/val_iceplains/0/000005.mp4,test/val_iceplains/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +72,72,262,test/val_iceplains/3/000006.mp4,test/val_iceplains/3/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +73,73,164,test/val_savanna/0/000015.mp4,test/val_savanna/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +74,74,337,test/val_iceplains/2/000036.mp4,test/val_iceplains/2/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +75,75,95,test/val_desert/0/000036.mp4,test/val_desert/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +76,76,277,test/val_iceplains/0/000013.mp4,test/val_iceplains/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +77,77,26,test/val_desert/1/000010.mp4,test/val_desert/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +78,78,77,test/val_desert/0/000029.mp4,test/val_desert/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +79,79,294,test/val_iceplains/1/000019.mp4,test/val_iceplains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +80,80,34,test/val_desert/1/000014.mp4,test/val_desert/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +81,81,188,test/val_savanna/1/000029.mp4,test/val_savanna/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +82,82,170,test/val_savanna/0/000017.mp4,test/val_savanna/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +83,83,415,test/val_plains/2/000017.mp4,test/val_plains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +84,84,145,test/val_savanna/0/000006.mp4,test/val_savanna/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +85,85,133,test/val_savanna/1/000002.mp4,test/val_savanna/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +86,86,347,test/val_iceplains/2/000039.mp4,test/val_iceplains/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +87,87,209,test/val_savanna/0/000038.mp4,test/val_savanna/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +88,88,404,test/val_plains/1/000013.mp4,test/val_plains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +89,89,211,test/val_savanna/2/000038.mp4,test/val_savanna/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +90,90,464,test/val_plains/2/000044.mp4,test/val_plains/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +91,91,162,test/val_savanna/1/000014.mp4,test/val_savanna/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +92,92,419,test/val_plains/2/000019.mp4,test/val_plains/2/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +93,93,127,test/val_desert/3/000049.mp4,test/val_desert/3/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +94,94,370,test/val_iceplains/2/000048.mp4,test/val_iceplains/2/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +95,95,293,test/val_iceplains/2/000018.mp4,test/val_iceplains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +96,96,251,test/val_iceplains/1/000003.mp4,test/val_iceplains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +97,97,441,test/val_plains/1/000031.mp4,test/val_plains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +98,98,69,test/val_desert/2/000026.mp4,test/val_desert/2/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +99,99,286,test/val_iceplains/2/000015.mp4,test/val_iceplains/2/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +100,100,438,test/val_plains/0/000029.mp4,test/val_plains/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +101,101,376,test/val_plains/0/000002.mp4,test/val_plains/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +102,102,163,test/val_savanna/2/000014.mp4,test/val_savanna/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +103,103,31,test/val_desert/1/000013.mp4,test/val_desert/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +104,104,241,test/val_savanna/0/000049.mp4,test/val_savanna/0/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +105,105,458,test/val_plains/0/000040.mp4,test/val_plains/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +106,106,371,test/val_iceplains/1/000049.mp4,test/val_iceplains/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +107,107,230,test/val_savanna/3/000045.mp4,test/val_savanna/3/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +108,108,372,test/val_iceplains/2/000049.mp4,test/val_iceplains/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +109,109,223,test/val_savanna/1/000042.mp4,test/val_savanna/1/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +110,110,68,test/val_desert/0/000026.mp4,test/val_desert/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +111,111,461,test/val_plains/2/000042.mp4,test/val_plains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +112,112,151,test/val_savanna/1/000009.mp4,test/val_savanna/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +113,113,72,test/val_desert/3/000027.mp4,test/val_desert/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +114,114,427,test/val_plains/0/000024.mp4,test/val_plains/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +115,115,317,test/val_iceplains/3/000029.mp4,test/val_iceplains/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +116,116,249,test/val_iceplains/2/000001.mp4,test/val_iceplains/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +117,117,364,test/val_iceplains/2/000046.mp4,test/val_iceplains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +118,118,96,test/val_desert/2/000037.mp4,test/val_desert/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +119,119,190,test/val_savanna/3/000029.mp4,test/val_savanna/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +120,120,227,test/val_savanna/2/000044.mp4,test/val_savanna/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +121,121,244,test/val_iceplains/0/000000.mp4,test/val_iceplains/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +122,122,38,test/val_desert/3/000015.mp4,test/val_desert/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +123,123,291,test/val_iceplains/2/000017.mp4,test/val_iceplains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +124,124,85,test/val_desert/2/000032.mp4,test/val_desert/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +125,125,358,test/val_iceplains/0/000044.mp4,test/val_iceplains/0/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +126,126,50,test/val_desert/1/000020.mp4,test/val_desert/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +127,127,207,test/val_savanna/0/000036.mp4,test/val_savanna/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +128,128,330,test/val_iceplains/2/000034.mp4,test/val_iceplains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +129,129,132,test/val_savanna/0/000001.mp4,test/val_savanna/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +130,130,423,test/val_plains/3/000021.mp4,test/val_plains/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +131,131,66,test/val_desert/2/000025.mp4,test/val_desert/2/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +132,132,153,test/val_savanna/3/000009.mp4,test/val_savanna/3/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +133,133,382,test/val_plains/0/000004.mp4,test/val_plains/0/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +134,134,143,test/val_savanna/2/000005.mp4,test/val_savanna/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +135,135,348,test/val_iceplains/1/000040.mp4,test/val_iceplains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +136,136,384,test/val_plains/2/000005.mp4,test/val_plains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +137,137,30,test/val_desert/0/000012.mp4,test/val_desert/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +138,138,203,test/val_savanna/2/000034.mp4,test/val_savanna/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +139,139,179,test/val_savanna/0/000022.mp4,test/val_savanna/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +140,140,253,test/val_iceplains/2/000003.mp4,test/val_iceplains/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +141,141,289,test/val_iceplains/2/000016.mp4,test/val_iceplains/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +142,142,422,test/val_plains/2/000020.mp4,test/val_plains/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +143,143,20,test/val_desert/0/000008.mp4,test/val_desert/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +144,144,130,test/val_savanna/0/000000.mp4,test/val_savanna/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +145,145,296,test/val_iceplains/1/000020.mp4,test/val_iceplains/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +146,146,278,test/val_iceplains/3/000013.mp4,test/val_iceplains/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +147,147,269,test/val_iceplains/0/000009.mp4,test/val_iceplains/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +148,148,314,test/val_iceplains/0/000028.mp4,test/val_iceplains/0/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +149,149,281,test/val_iceplains/0/000014.mp4,test/val_iceplains/0/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +150,150,149,test/val_savanna/0/000008.mp4,test/val_savanna/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +151,151,184,test/val_savanna/0/000024.mp4,test/val_savanna/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +152,152,407,test/val_plains/1/000014.mp4,test/val_plains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +153,153,304,test/val_iceplains/2/000023.mp4,test/val_iceplains/2/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +154,154,208,test/val_savanna/3/000037.mp4,test/val_savanna/3/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +155,155,182,test/val_savanna/3/000023.mp4,test/val_savanna/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +156,156,394,test/val_plains/1/000009.mp4,test/val_plains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +157,157,14,test/val_desert/0/000006.mp4,test/val_desert/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +158,158,55,test/val_desert/3/000021.mp4,test/val_desert/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +159,159,75,test/val_desert/2/000028.mp4,test/val_desert/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +160,160,319,test/val_iceplains/1/000031.mp4,test/val_iceplains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +161,161,39,test/val_desert/1/000016.mp4,test/val_desert/1/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +162,162,418,test/val_plains/1/000019.mp4,test/val_plains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +163,163,389,test/val_plains/3/000007.mp4,test/val_plains/3/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +164,164,409,test/val_plains/2/000014.mp4,test/val_plains/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +165,165,344,test/val_iceplains/2/000038.mp4,test/val_iceplains/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +166,166,396,test/val_plains/2/000009.mp4,test/val_plains/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +167,167,67,test/val_desert/1/000026.mp4,test/val_desert/1/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +168,168,99,test/val_desert/3/000038.mp4,test/val_desert/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +169,169,288,test/val_iceplains/3/000016.mp4,test/val_iceplains/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +170,170,97,test/val_desert/1/000038.mp4,test/val_desert/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +171,171,178,test/val_savanna/2/000021.mp4,test/val_savanna/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +172,172,264,test/val_iceplains/1/000007.mp4,test/val_iceplains/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +173,173,266,test/val_iceplains/0/000008.mp4,test/val_iceplains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +174,174,303,test/val_iceplains/1/000023.mp4,test/val_iceplains/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +175,175,442,test/val_plains/0/000031.mp4,test/val_plains/0/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +176,176,233,test/val_savanna/1/000047.mp4,test/val_savanna/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +177,177,240,test/val_savanna/1/000049.mp4,test/val_savanna/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +178,178,42,test/val_desert/0/000017.mp4,test/val_desert/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +179,179,369,test/val_iceplains/3/000048.mp4,test/val_iceplains/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +180,180,398,test/val_plains/3/000010.mp4,test/val_plains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +181,181,435,test/val_plains/2/000027.mp4,test/val_plains/2/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +182,182,242,test/val_savanna/2/000049.mp4,test/val_savanna/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +183,183,361,test/val_iceplains/2/000045.mp4,test/val_iceplains/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +184,184,470,test/val_plains/2/000047.mp4,test/val_plains/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +185,185,356,test/val_iceplains/2/000042.mp4,test/val_iceplains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +186,186,468,test/val_plains/2/000046.mp4,test/val_plains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +187,187,113,test/val_desert/0/000043.mp4,test/val_desert/0/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +188,188,19,test/val_desert/1/000008.mp4,test/val_desert/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +189,189,199,test/val_savanna/0/000033.mp4,test/val_savanna/0/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +190,190,118,test/val_desert/2/000045.mp4,test/val_desert/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +191,191,433,test/val_plains/0/000027.mp4,test/val_plains/0/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +192,192,328,test/val_iceplains/0/000034.mp4,test/val_iceplains/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +193,193,213,test/val_savanna/0/000039.mp4,test/val_savanna/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +194,194,339,test/val_iceplains/0/000037.mp4,test/val_iceplains/0/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +195,195,466,test/val_plains/1/000046.mp4,test/val_plains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +196,196,355,test/val_iceplains/3/000042.mp4,test/val_iceplains/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +197,197,336,test/val_iceplains/3/000036.mp4,test/val_iceplains/3/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +198,198,420,test/val_plains/0/000020.mp4,test/val_plains/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +199,199,410,test/val_plains/1/000015.mp4,test/val_plains/1/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +200,200,27,test/val_desert/1/000011.mp4,test/val_desert/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +201,201,86,test/val_desert/1/000033.mp4,test/val_desert/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +202,202,210,test/val_savanna/3/000038.mp4,test/val_savanna/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +203,203,103,test/val_desert/3/000039.mp4,test/val_desert/3/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +204,204,245,test/val_iceplains/3/000000.mp4,test/val_iceplains/3/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +205,205,25,test/val_desert/2/000009.mp4,test/val_desert/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +206,206,340,test/val_iceplains/2/000037.mp4,test/val_iceplains/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +207,207,167,test/val_savanna/0/000016.mp4,test/val_savanna/0/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +208,208,424,test/val_plains/0/000022.mp4,test/val_plains/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +209,209,448,test/val_plains/2/000033.mp4,test/val_plains/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +210,210,386,test/val_plains/0/000006.mp4,test/val_plains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +211,211,392,test/val_plains/3/000008.mp4,test/val_plains/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +212,212,105,test/val_desert/1/000040.mp4,test/val_desert/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +213,213,267,test/val_iceplains/2/000008.mp4,test/val_iceplains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +214,214,90,test/val_desert/0/000034.mp4,test/val_desert/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +215,215,426,test/val_plains/0/000023.mp4,test/val_plains/0/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +216,216,168,test/val_savanna/3/000016.mp4,test/val_savanna/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +217,217,218,test/val_savanna/3/000040.mp4,test/val_savanna/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +218,218,445,test/val_plains/0/000032.mp4,test/val_plains/0/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +219,219,48,test/val_desert/3/000019.mp4,test/val_desert/3/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +220,220,104,test/val_desert/2/000039.mp4,test/val_desert/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +221,221,206,test/val_savanna/2/000035.mp4,test/val_savanna/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +222,222,228,test/val_savanna/1/000045.mp4,test/val_savanna/1/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +223,223,200,test/val_savanna/2/000033.mp4,test/val_savanna/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +224,224,62,test/val_desert/2/000024.mp4,test/val_desert/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +225,225,121,test/val_desert/1/000047.mp4,test/val_desert/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +226,226,147,test/val_savanna/0/000007.mp4,test/val_savanna/0/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +227,227,146,test/val_savanna/2/000006.mp4,test/val_savanna/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +228,228,225,test/val_savanna/1/000043.mp4,test/val_savanna/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +229,229,313,test/val_iceplains/3/000027.mp4,test/val_iceplains/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +230,230,354,test/val_iceplains/0/000042.mp4,test/val_iceplains/0/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +231,231,367,test/val_iceplains/1/000048.mp4,test/val_iceplains/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +232,232,297,test/val_iceplains/3/000020.mp4,test/val_iceplains/3/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +233,233,124,test/val_desert/1/000048.mp4,test/val_desert/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +234,234,171,test/val_savanna/2/000017.mp4,test/val_savanna/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +235,235,342,test/val_iceplains/0/000038.mp4,test/val_iceplains/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +236,236,60,test/val_desert/3/000023.mp4,test/val_desert/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +237,237,131,test/val_savanna/1/000001.mp4,test/val_savanna/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +238,238,365,test/val_iceplains/1/000047.mp4,test/val_iceplains/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +239,239,315,test/val_iceplains/2/000028.mp4,test/val_iceplains/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +240,240,220,test/val_savanna/1/000041.mp4,test/val_savanna/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +241,241,8,test/val_desert/1/000003.mp4,test/val_desert/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +242,242,261,test/val_iceplains/0/000006.mp4,test/val_iceplains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +243,243,17,test/val_desert/1/000007.mp4,test/val_desert/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +244,244,462,test/val_plains/1/000043.mp4,test/val_plains/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +245,245,454,test/val_plains/3/000038.mp4,test/val_plains/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +246,246,351,test/val_iceplains/1/000041.mp4,test/val_iceplains/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +247,247,383,test/val_plains/3/000005.mp4,test/val_plains/3/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +248,248,18,test/val_desert/2/000007.mp4,test/val_desert/2/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +249,249,457,test/val_plains/1/000040.mp4,test/val_plains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +250,250,326,test/val_iceplains/3/000033.mp4,test/val_iceplains/3/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +251,251,254,test/val_iceplains/1/000004.mp4,test/val_iceplains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +252,252,375,test/val_plains/1/000002.mp4,test/val_plains/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +253,253,141,test/val_savanna/0/000005.mp4,test/val_savanna/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +254,254,306,test/val_iceplains/2/000024.mp4,test/val_iceplains/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +255,255,10,test/val_desert/2/000003.mp4,test/val_desert/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +256,256,173,test/val_savanna/0/000020.mp4,test/val_savanna/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +257,257,320,test/val_iceplains/1/000032.mp4,test/val_iceplains/1/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +258,258,115,test/val_desert/1/000044.mp4,test/val_desert/1/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +259,259,341,test/val_iceplains/1/000038.mp4,test/val_iceplains/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +260,260,61,test/val_desert/3/000024.mp4,test/val_desert/3/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +261,261,231,test/val_savanna/1/000046.mp4,test/val_savanna/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +262,262,224,test/val_savanna/3/000042.mp4,test/val_savanna/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +263,263,292,test/val_iceplains/3/000018.mp4,test/val_iceplains/3/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +264,264,334,test/val_iceplains/1/000036.mp4,test/val_iceplains/1/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +265,265,159,test/val_savanna/0/000013.mp4,test/val_savanna/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +266,266,102,test/val_desert/0/000039.mp4,test/val_desert/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +267,267,362,test/val_iceplains/1/000046.mp4,test/val_iceplains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +268,268,322,test/val_iceplains/3/000032.mp4,test/val_iceplains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +269,269,381,test/val_plains/1/000004.mp4,test/val_plains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +270,270,84,test/val_desert/3/000032.mp4,test/val_desert/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +271,271,243,test/val_iceplains/1/000000.mp4,test/val_iceplains/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +272,272,65,test/val_desert/3/000025.mp4,test/val_desert/3/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +273,273,64,test/val_desert/0/000025.mp4,test/val_desert/0/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +274,274,272,test/val_iceplains/2/000010.mp4,test/val_iceplains/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +275,275,399,test/val_plains/0/000011.mp4,test/val_plains/0/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +276,276,248,test/val_iceplains/3/000001.mp4,test/val_iceplains/3/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +277,277,43,test/val_desert/3/000017.mp4,test/val_desert/3/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +278,278,393,test/val_plains/2/000008.mp4,test/val_plains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +279,279,123,test/val_desert/2/000047.mp4,test/val_desert/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +280,280,29,test/val_desert/2/000011.mp4,test/val_desert/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +281,281,122,test/val_desert/0/000047.mp4,test/val_desert/0/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +282,282,157,test/val_savanna/3/000011.mp4,test/val_savanna/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +283,283,333,test/val_iceplains/2/000035.mp4,test/val_iceplains/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +284,284,307,test/val_iceplains/1/000025.mp4,test/val_iceplains/1/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +285,285,9,test/val_desert/0/000003.mp4,test/val_desert/0/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +286,286,250,test/val_iceplains/2/000002.mp4,test/val_iceplains/2/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +287,287,165,test/val_savanna/3/000015.mp4,test/val_savanna/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +288,288,0,test/val_desert/1/000000.mp4,test/val_desert/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +289,289,271,test/val_iceplains/3/000010.mp4,test/val_iceplains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +290,290,276,test/val_iceplains/1/000013.mp4,test/val_iceplains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +291,291,82,test/val_desert/1/000031.mp4,test/val_desert/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +292,292,191,test/val_savanna/2/000029.mp4,test/val_savanna/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +293,293,374,test/val_plains/1/000001.mp4,test/val_plains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +294,294,446,test/val_plains/3/000032.mp4,test/val_plains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +295,295,378,test/val_plains/1/000003.mp4,test/val_plains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +296,296,152,test/val_savanna/0/000009.mp4,test/val_savanna/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +297,297,54,test/val_desert/0/000021.mp4,test/val_desert/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +298,298,259,test/val_iceplains/2/000005.mp4,test/val_iceplains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +299,299,91,test/val_desert/2/000034.mp4,test/val_desert/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.sha256 b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..065350069e5b0f46e33cd51ad63f60aa956822fd --- /dev/null +++ b/outputs/dememwm_long_horizon/dememwm_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.sha256 @@ -0,0 +1 @@ +ef4e4de9d3975d8a7e724bb609e57f20186cd1f7285457c00fb85c1aca973b99 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/evaluation.patch b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/evaluation.patch new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..3cfd08bc49f4159d2bd5cace6409ac92956d78e1 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.csv @@ -0,0 +1,51 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.sha256 b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..67b0dc2042229c0768d324282215bdf64bebe28f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.sha256 @@ -0,0 +1 @@ +f9536911e434610cce23306ff99575db53cd7fc29e3df841da4b6631bf418634 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/git_provenance.json b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/git_provenance.json new file mode 100644 index 0000000000000000000000000000000000000000..fabe691fc08806f957f2445d4a52b2f38d7f07d8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/git_provenance.json @@ -0,0 +1,5 @@ +{ + "branch": "dememwm-unified-sourcefree-age-balanced", + "head": "732555281e86c4fdbb71c98a000d648c3edc3109", + "status": "?? results/\n?? train_dememwm_full_berzelius_gate_off_h200.sh\n?? train_dememwm_separate_mem_gatefix_4layers512_valid500_berzelius_h200.sh" +} diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..dae70b4071c2d2a834268c2a26f061d308f11f2c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.csv @@ -0,0 +1,301 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +50,50,212,test/val_savanna/1/000039.mp4,test/val_savanna/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +51,51,4,test/val_desert/0/000001.mp4,test/val_desert/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +52,52,7,test/val_desert/0/000002.mp4,test/val_desert/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +53,53,310,test/val_iceplains/0/000026.mp4,test/val_iceplains/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +54,54,284,test/val_iceplains/0/000015.mp4,test/val_iceplains/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +55,55,109,test/val_desert/3/000041.mp4,test/val_desert/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +56,56,45,test/val_desert/0/000018.mp4,test/val_desert/0/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +57,57,93,test/val_desert/0/000035.mp4,test/val_desert/0/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +58,58,400,test/val_plains/3/000011.mp4,test/val_plains/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +59,59,300,test/val_iceplains/2/000021.mp4,test/val_iceplains/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +60,60,226,test/val_savanna/2/000043.mp4,test/val_savanna/2/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +61,61,417,test/val_plains/2/000018.mp4,test/val_plains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +62,62,120,test/val_desert/3/000046.mp4,test/val_desert/3/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +63,63,106,test/val_desert/3/000040.mp4,test/val_desert/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +64,64,345,test/val_iceplains/1/000039.mp4,test/val_iceplains/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +65,65,169,test/val_savanna/2/000016.mp4,test/val_savanna/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +66,66,312,test/val_iceplains/1/000027.mp4,test/val_iceplains/1/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +67,67,368,test/val_iceplains/0/000048.mp4,test/val_iceplains/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +68,68,449,test/val_plains/1/000034.mp4,test/val_plains/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +69,69,11,test/val_desert/1/000004.mp4,test/val_desert/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +70,70,6,test/val_desert/2/000001.mp4,test/val_desert/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +71,71,257,test/val_iceplains/0/000005.mp4,test/val_iceplains/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +72,72,262,test/val_iceplains/3/000006.mp4,test/val_iceplains/3/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +73,73,164,test/val_savanna/0/000015.mp4,test/val_savanna/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +74,74,337,test/val_iceplains/2/000036.mp4,test/val_iceplains/2/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +75,75,95,test/val_desert/0/000036.mp4,test/val_desert/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +76,76,277,test/val_iceplains/0/000013.mp4,test/val_iceplains/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +77,77,26,test/val_desert/1/000010.mp4,test/val_desert/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +78,78,77,test/val_desert/0/000029.mp4,test/val_desert/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +79,79,294,test/val_iceplains/1/000019.mp4,test/val_iceplains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +80,80,34,test/val_desert/1/000014.mp4,test/val_desert/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +81,81,188,test/val_savanna/1/000029.mp4,test/val_savanna/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +82,82,170,test/val_savanna/0/000017.mp4,test/val_savanna/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +83,83,415,test/val_plains/2/000017.mp4,test/val_plains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +84,84,145,test/val_savanna/0/000006.mp4,test/val_savanna/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +85,85,133,test/val_savanna/1/000002.mp4,test/val_savanna/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +86,86,347,test/val_iceplains/2/000039.mp4,test/val_iceplains/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +87,87,209,test/val_savanna/0/000038.mp4,test/val_savanna/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +88,88,404,test/val_plains/1/000013.mp4,test/val_plains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +89,89,211,test/val_savanna/2/000038.mp4,test/val_savanna/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +90,90,464,test/val_plains/2/000044.mp4,test/val_plains/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +91,91,162,test/val_savanna/1/000014.mp4,test/val_savanna/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +92,92,419,test/val_plains/2/000019.mp4,test/val_plains/2/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +93,93,127,test/val_desert/3/000049.mp4,test/val_desert/3/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +94,94,370,test/val_iceplains/2/000048.mp4,test/val_iceplains/2/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +95,95,293,test/val_iceplains/2/000018.mp4,test/val_iceplains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +96,96,251,test/val_iceplains/1/000003.mp4,test/val_iceplains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +97,97,441,test/val_plains/1/000031.mp4,test/val_plains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +98,98,69,test/val_desert/2/000026.mp4,test/val_desert/2/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +99,99,286,test/val_iceplains/2/000015.mp4,test/val_iceplains/2/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +100,100,438,test/val_plains/0/000029.mp4,test/val_plains/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +101,101,376,test/val_plains/0/000002.mp4,test/val_plains/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +102,102,163,test/val_savanna/2/000014.mp4,test/val_savanna/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +103,103,31,test/val_desert/1/000013.mp4,test/val_desert/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +104,104,241,test/val_savanna/0/000049.mp4,test/val_savanna/0/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +105,105,458,test/val_plains/0/000040.mp4,test/val_plains/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +106,106,371,test/val_iceplains/1/000049.mp4,test/val_iceplains/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +107,107,230,test/val_savanna/3/000045.mp4,test/val_savanna/3/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +108,108,372,test/val_iceplains/2/000049.mp4,test/val_iceplains/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +109,109,223,test/val_savanna/1/000042.mp4,test/val_savanna/1/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +110,110,68,test/val_desert/0/000026.mp4,test/val_desert/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +111,111,461,test/val_plains/2/000042.mp4,test/val_plains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +112,112,151,test/val_savanna/1/000009.mp4,test/val_savanna/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +113,113,72,test/val_desert/3/000027.mp4,test/val_desert/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +114,114,427,test/val_plains/0/000024.mp4,test/val_plains/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +115,115,317,test/val_iceplains/3/000029.mp4,test/val_iceplains/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +116,116,249,test/val_iceplains/2/000001.mp4,test/val_iceplains/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +117,117,364,test/val_iceplains/2/000046.mp4,test/val_iceplains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +118,118,96,test/val_desert/2/000037.mp4,test/val_desert/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +119,119,190,test/val_savanna/3/000029.mp4,test/val_savanna/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +120,120,227,test/val_savanna/2/000044.mp4,test/val_savanna/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +121,121,244,test/val_iceplains/0/000000.mp4,test/val_iceplains/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +122,122,38,test/val_desert/3/000015.mp4,test/val_desert/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +123,123,291,test/val_iceplains/2/000017.mp4,test/val_iceplains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +124,124,85,test/val_desert/2/000032.mp4,test/val_desert/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +125,125,358,test/val_iceplains/0/000044.mp4,test/val_iceplains/0/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +126,126,50,test/val_desert/1/000020.mp4,test/val_desert/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +127,127,207,test/val_savanna/0/000036.mp4,test/val_savanna/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +128,128,330,test/val_iceplains/2/000034.mp4,test/val_iceplains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +129,129,132,test/val_savanna/0/000001.mp4,test/val_savanna/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +130,130,423,test/val_plains/3/000021.mp4,test/val_plains/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +131,131,66,test/val_desert/2/000025.mp4,test/val_desert/2/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +132,132,153,test/val_savanna/3/000009.mp4,test/val_savanna/3/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +133,133,382,test/val_plains/0/000004.mp4,test/val_plains/0/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +134,134,143,test/val_savanna/2/000005.mp4,test/val_savanna/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +135,135,348,test/val_iceplains/1/000040.mp4,test/val_iceplains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +136,136,384,test/val_plains/2/000005.mp4,test/val_plains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +137,137,30,test/val_desert/0/000012.mp4,test/val_desert/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +138,138,203,test/val_savanna/2/000034.mp4,test/val_savanna/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +139,139,179,test/val_savanna/0/000022.mp4,test/val_savanna/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +140,140,253,test/val_iceplains/2/000003.mp4,test/val_iceplains/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +141,141,289,test/val_iceplains/2/000016.mp4,test/val_iceplains/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +142,142,422,test/val_plains/2/000020.mp4,test/val_plains/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +143,143,20,test/val_desert/0/000008.mp4,test/val_desert/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +144,144,130,test/val_savanna/0/000000.mp4,test/val_savanna/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +145,145,296,test/val_iceplains/1/000020.mp4,test/val_iceplains/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +146,146,278,test/val_iceplains/3/000013.mp4,test/val_iceplains/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +147,147,269,test/val_iceplains/0/000009.mp4,test/val_iceplains/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +148,148,314,test/val_iceplains/0/000028.mp4,test/val_iceplains/0/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +149,149,281,test/val_iceplains/0/000014.mp4,test/val_iceplains/0/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +150,150,149,test/val_savanna/0/000008.mp4,test/val_savanna/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +151,151,184,test/val_savanna/0/000024.mp4,test/val_savanna/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +152,152,407,test/val_plains/1/000014.mp4,test/val_plains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +153,153,304,test/val_iceplains/2/000023.mp4,test/val_iceplains/2/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +154,154,208,test/val_savanna/3/000037.mp4,test/val_savanna/3/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +155,155,182,test/val_savanna/3/000023.mp4,test/val_savanna/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +156,156,394,test/val_plains/1/000009.mp4,test/val_plains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +157,157,14,test/val_desert/0/000006.mp4,test/val_desert/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +158,158,55,test/val_desert/3/000021.mp4,test/val_desert/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +159,159,75,test/val_desert/2/000028.mp4,test/val_desert/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +160,160,319,test/val_iceplains/1/000031.mp4,test/val_iceplains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +161,161,39,test/val_desert/1/000016.mp4,test/val_desert/1/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +162,162,418,test/val_plains/1/000019.mp4,test/val_plains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +163,163,389,test/val_plains/3/000007.mp4,test/val_plains/3/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +164,164,409,test/val_plains/2/000014.mp4,test/val_plains/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +165,165,344,test/val_iceplains/2/000038.mp4,test/val_iceplains/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +166,166,396,test/val_plains/2/000009.mp4,test/val_plains/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +167,167,67,test/val_desert/1/000026.mp4,test/val_desert/1/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +168,168,99,test/val_desert/3/000038.mp4,test/val_desert/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +169,169,288,test/val_iceplains/3/000016.mp4,test/val_iceplains/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +170,170,97,test/val_desert/1/000038.mp4,test/val_desert/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +171,171,178,test/val_savanna/2/000021.mp4,test/val_savanna/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +172,172,264,test/val_iceplains/1/000007.mp4,test/val_iceplains/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +173,173,266,test/val_iceplains/0/000008.mp4,test/val_iceplains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +174,174,303,test/val_iceplains/1/000023.mp4,test/val_iceplains/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +175,175,442,test/val_plains/0/000031.mp4,test/val_plains/0/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +176,176,233,test/val_savanna/1/000047.mp4,test/val_savanna/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +177,177,240,test/val_savanna/1/000049.mp4,test/val_savanna/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +178,178,42,test/val_desert/0/000017.mp4,test/val_desert/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +179,179,369,test/val_iceplains/3/000048.mp4,test/val_iceplains/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +180,180,398,test/val_plains/3/000010.mp4,test/val_plains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +181,181,435,test/val_plains/2/000027.mp4,test/val_plains/2/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +182,182,242,test/val_savanna/2/000049.mp4,test/val_savanna/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +183,183,361,test/val_iceplains/2/000045.mp4,test/val_iceplains/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +184,184,470,test/val_plains/2/000047.mp4,test/val_plains/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +185,185,356,test/val_iceplains/2/000042.mp4,test/val_iceplains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +186,186,468,test/val_plains/2/000046.mp4,test/val_plains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +187,187,113,test/val_desert/0/000043.mp4,test/val_desert/0/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +188,188,19,test/val_desert/1/000008.mp4,test/val_desert/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +189,189,199,test/val_savanna/0/000033.mp4,test/val_savanna/0/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +190,190,118,test/val_desert/2/000045.mp4,test/val_desert/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +191,191,433,test/val_plains/0/000027.mp4,test/val_plains/0/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +192,192,328,test/val_iceplains/0/000034.mp4,test/val_iceplains/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +193,193,213,test/val_savanna/0/000039.mp4,test/val_savanna/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +194,194,339,test/val_iceplains/0/000037.mp4,test/val_iceplains/0/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +195,195,466,test/val_plains/1/000046.mp4,test/val_plains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +196,196,355,test/val_iceplains/3/000042.mp4,test/val_iceplains/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +197,197,336,test/val_iceplains/3/000036.mp4,test/val_iceplains/3/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +198,198,420,test/val_plains/0/000020.mp4,test/val_plains/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +199,199,410,test/val_plains/1/000015.mp4,test/val_plains/1/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +200,200,27,test/val_desert/1/000011.mp4,test/val_desert/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +201,201,86,test/val_desert/1/000033.mp4,test/val_desert/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +202,202,210,test/val_savanna/3/000038.mp4,test/val_savanna/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +203,203,103,test/val_desert/3/000039.mp4,test/val_desert/3/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +204,204,245,test/val_iceplains/3/000000.mp4,test/val_iceplains/3/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +205,205,25,test/val_desert/2/000009.mp4,test/val_desert/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +206,206,340,test/val_iceplains/2/000037.mp4,test/val_iceplains/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +207,207,167,test/val_savanna/0/000016.mp4,test/val_savanna/0/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +208,208,424,test/val_plains/0/000022.mp4,test/val_plains/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +209,209,448,test/val_plains/2/000033.mp4,test/val_plains/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +210,210,386,test/val_plains/0/000006.mp4,test/val_plains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +211,211,392,test/val_plains/3/000008.mp4,test/val_plains/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +212,212,105,test/val_desert/1/000040.mp4,test/val_desert/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +213,213,267,test/val_iceplains/2/000008.mp4,test/val_iceplains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +214,214,90,test/val_desert/0/000034.mp4,test/val_desert/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +215,215,426,test/val_plains/0/000023.mp4,test/val_plains/0/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +216,216,168,test/val_savanna/3/000016.mp4,test/val_savanna/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +217,217,218,test/val_savanna/3/000040.mp4,test/val_savanna/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +218,218,445,test/val_plains/0/000032.mp4,test/val_plains/0/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +219,219,48,test/val_desert/3/000019.mp4,test/val_desert/3/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +220,220,104,test/val_desert/2/000039.mp4,test/val_desert/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +221,221,206,test/val_savanna/2/000035.mp4,test/val_savanna/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +222,222,228,test/val_savanna/1/000045.mp4,test/val_savanna/1/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +223,223,200,test/val_savanna/2/000033.mp4,test/val_savanna/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +224,224,62,test/val_desert/2/000024.mp4,test/val_desert/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +225,225,121,test/val_desert/1/000047.mp4,test/val_desert/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +226,226,147,test/val_savanna/0/000007.mp4,test/val_savanna/0/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +227,227,146,test/val_savanna/2/000006.mp4,test/val_savanna/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +228,228,225,test/val_savanna/1/000043.mp4,test/val_savanna/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +229,229,313,test/val_iceplains/3/000027.mp4,test/val_iceplains/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +230,230,354,test/val_iceplains/0/000042.mp4,test/val_iceplains/0/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +231,231,367,test/val_iceplains/1/000048.mp4,test/val_iceplains/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +232,232,297,test/val_iceplains/3/000020.mp4,test/val_iceplains/3/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +233,233,124,test/val_desert/1/000048.mp4,test/val_desert/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +234,234,171,test/val_savanna/2/000017.mp4,test/val_savanna/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +235,235,342,test/val_iceplains/0/000038.mp4,test/val_iceplains/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +236,236,60,test/val_desert/3/000023.mp4,test/val_desert/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +237,237,131,test/val_savanna/1/000001.mp4,test/val_savanna/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +238,238,365,test/val_iceplains/1/000047.mp4,test/val_iceplains/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +239,239,315,test/val_iceplains/2/000028.mp4,test/val_iceplains/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +240,240,220,test/val_savanna/1/000041.mp4,test/val_savanna/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +241,241,8,test/val_desert/1/000003.mp4,test/val_desert/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +242,242,261,test/val_iceplains/0/000006.mp4,test/val_iceplains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +243,243,17,test/val_desert/1/000007.mp4,test/val_desert/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +244,244,462,test/val_plains/1/000043.mp4,test/val_plains/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +245,245,454,test/val_plains/3/000038.mp4,test/val_plains/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +246,246,351,test/val_iceplains/1/000041.mp4,test/val_iceplains/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +247,247,383,test/val_plains/3/000005.mp4,test/val_plains/3/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +248,248,18,test/val_desert/2/000007.mp4,test/val_desert/2/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +249,249,457,test/val_plains/1/000040.mp4,test/val_plains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +250,250,326,test/val_iceplains/3/000033.mp4,test/val_iceplains/3/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +251,251,254,test/val_iceplains/1/000004.mp4,test/val_iceplains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +252,252,375,test/val_plains/1/000002.mp4,test/val_plains/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +253,253,141,test/val_savanna/0/000005.mp4,test/val_savanna/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +254,254,306,test/val_iceplains/2/000024.mp4,test/val_iceplains/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +255,255,10,test/val_desert/2/000003.mp4,test/val_desert/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +256,256,173,test/val_savanna/0/000020.mp4,test/val_savanna/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +257,257,320,test/val_iceplains/1/000032.mp4,test/val_iceplains/1/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +258,258,115,test/val_desert/1/000044.mp4,test/val_desert/1/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +259,259,341,test/val_iceplains/1/000038.mp4,test/val_iceplains/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +260,260,61,test/val_desert/3/000024.mp4,test/val_desert/3/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +261,261,231,test/val_savanna/1/000046.mp4,test/val_savanna/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +262,262,224,test/val_savanna/3/000042.mp4,test/val_savanna/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +263,263,292,test/val_iceplains/3/000018.mp4,test/val_iceplains/3/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +264,264,334,test/val_iceplains/1/000036.mp4,test/val_iceplains/1/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +265,265,159,test/val_savanna/0/000013.mp4,test/val_savanna/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +266,266,102,test/val_desert/0/000039.mp4,test/val_desert/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +267,267,362,test/val_iceplains/1/000046.mp4,test/val_iceplains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +268,268,322,test/val_iceplains/3/000032.mp4,test/val_iceplains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +269,269,381,test/val_plains/1/000004.mp4,test/val_plains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +270,270,84,test/val_desert/3/000032.mp4,test/val_desert/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +271,271,243,test/val_iceplains/1/000000.mp4,test/val_iceplains/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +272,272,65,test/val_desert/3/000025.mp4,test/val_desert/3/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +273,273,64,test/val_desert/0/000025.mp4,test/val_desert/0/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +274,274,272,test/val_iceplains/2/000010.mp4,test/val_iceplains/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +275,275,399,test/val_plains/0/000011.mp4,test/val_plains/0/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +276,276,248,test/val_iceplains/3/000001.mp4,test/val_iceplains/3/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +277,277,43,test/val_desert/3/000017.mp4,test/val_desert/3/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +278,278,393,test/val_plains/2/000008.mp4,test/val_plains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +279,279,123,test/val_desert/2/000047.mp4,test/val_desert/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +280,280,29,test/val_desert/2/000011.mp4,test/val_desert/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +281,281,122,test/val_desert/0/000047.mp4,test/val_desert/0/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +282,282,157,test/val_savanna/3/000011.mp4,test/val_savanna/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +283,283,333,test/val_iceplains/2/000035.mp4,test/val_iceplains/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +284,284,307,test/val_iceplains/1/000025.mp4,test/val_iceplains/1/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +285,285,9,test/val_desert/0/000003.mp4,test/val_desert/0/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +286,286,250,test/val_iceplains/2/000002.mp4,test/val_iceplains/2/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +287,287,165,test/val_savanna/3/000015.mp4,test/val_savanna/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +288,288,0,test/val_desert/1/000000.mp4,test/val_desert/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +289,289,271,test/val_iceplains/3/000010.mp4,test/val_iceplains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +290,290,276,test/val_iceplains/1/000013.mp4,test/val_iceplains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +291,291,82,test/val_desert/1/000031.mp4,test/val_desert/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +292,292,191,test/val_savanna/2/000029.mp4,test/val_savanna/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +293,293,374,test/val_plains/1/000001.mp4,test/val_plains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +294,294,446,test/val_plains/3/000032.mp4,test/val_plains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +295,295,378,test/val_plains/1/000003.mp4,test/val_plains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +296,296,152,test/val_savanna/0/000009.mp4,test/val_savanna/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +297,297,54,test/val_desert/0/000021.mp4,test/val_desert/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +298,298,259,test/val_iceplains/2/000005.mp4,test/val_iceplains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +299,299,91,test/val_desert/2/000034.mp4,test/val_desert/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.sha256 b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..065350069e5b0f46e33cd51ad63f60aa956822fd --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.sha256 @@ -0,0 +1 @@ +ef4e4de9d3975d8a7e724bb609e57f20186cd1f7285457c00fb85c1aca973b99 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/evaluation.patch b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/evaluation.patch new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..3cfd08bc49f4159d2bd5cace6409ac92956d78e1 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.csv @@ -0,0 +1,51 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.sha256 b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..67b0dc2042229c0768d324282215bdf64bebe28f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.sha256 @@ -0,0 +1 @@ +f9536911e434610cce23306ff99575db53cd7fc29e3df841da4b6631bf418634 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/git_provenance.json b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/git_provenance.json new file mode 100644 index 0000000000000000000000000000000000000000..fabe691fc08806f957f2445d4a52b2f38d7f07d8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/git_provenance.json @@ -0,0 +1,5 @@ +{ + "branch": "dememwm-unified-sourcefree-age-balanced", + "head": "732555281e86c4fdbb71c98a000d648c3edc3109", + "status": "?? results/\n?? train_dememwm_full_berzelius_gate_off_h200.sh\n?? train_dememwm_separate_mem_gatefix_4layers512_valid500_berzelius_h200.sh" +} diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..dae70b4071c2d2a834268c2a26f061d308f11f2c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.csv @@ -0,0 +1,301 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +50,50,212,test/val_savanna/1/000039.mp4,test/val_savanna/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +51,51,4,test/val_desert/0/000001.mp4,test/val_desert/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +52,52,7,test/val_desert/0/000002.mp4,test/val_desert/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +53,53,310,test/val_iceplains/0/000026.mp4,test/val_iceplains/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +54,54,284,test/val_iceplains/0/000015.mp4,test/val_iceplains/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +55,55,109,test/val_desert/3/000041.mp4,test/val_desert/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +56,56,45,test/val_desert/0/000018.mp4,test/val_desert/0/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +57,57,93,test/val_desert/0/000035.mp4,test/val_desert/0/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +58,58,400,test/val_plains/3/000011.mp4,test/val_plains/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +59,59,300,test/val_iceplains/2/000021.mp4,test/val_iceplains/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +60,60,226,test/val_savanna/2/000043.mp4,test/val_savanna/2/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +61,61,417,test/val_plains/2/000018.mp4,test/val_plains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +62,62,120,test/val_desert/3/000046.mp4,test/val_desert/3/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +63,63,106,test/val_desert/3/000040.mp4,test/val_desert/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +64,64,345,test/val_iceplains/1/000039.mp4,test/val_iceplains/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +65,65,169,test/val_savanna/2/000016.mp4,test/val_savanna/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +66,66,312,test/val_iceplains/1/000027.mp4,test/val_iceplains/1/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +67,67,368,test/val_iceplains/0/000048.mp4,test/val_iceplains/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +68,68,449,test/val_plains/1/000034.mp4,test/val_plains/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +69,69,11,test/val_desert/1/000004.mp4,test/val_desert/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +70,70,6,test/val_desert/2/000001.mp4,test/val_desert/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +71,71,257,test/val_iceplains/0/000005.mp4,test/val_iceplains/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +72,72,262,test/val_iceplains/3/000006.mp4,test/val_iceplains/3/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +73,73,164,test/val_savanna/0/000015.mp4,test/val_savanna/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +74,74,337,test/val_iceplains/2/000036.mp4,test/val_iceplains/2/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +75,75,95,test/val_desert/0/000036.mp4,test/val_desert/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +76,76,277,test/val_iceplains/0/000013.mp4,test/val_iceplains/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +77,77,26,test/val_desert/1/000010.mp4,test/val_desert/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +78,78,77,test/val_desert/0/000029.mp4,test/val_desert/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +79,79,294,test/val_iceplains/1/000019.mp4,test/val_iceplains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +80,80,34,test/val_desert/1/000014.mp4,test/val_desert/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +81,81,188,test/val_savanna/1/000029.mp4,test/val_savanna/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +82,82,170,test/val_savanna/0/000017.mp4,test/val_savanna/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +83,83,415,test/val_plains/2/000017.mp4,test/val_plains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +84,84,145,test/val_savanna/0/000006.mp4,test/val_savanna/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +85,85,133,test/val_savanna/1/000002.mp4,test/val_savanna/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +86,86,347,test/val_iceplains/2/000039.mp4,test/val_iceplains/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +87,87,209,test/val_savanna/0/000038.mp4,test/val_savanna/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +88,88,404,test/val_plains/1/000013.mp4,test/val_plains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +89,89,211,test/val_savanna/2/000038.mp4,test/val_savanna/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +90,90,464,test/val_plains/2/000044.mp4,test/val_plains/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +91,91,162,test/val_savanna/1/000014.mp4,test/val_savanna/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +92,92,419,test/val_plains/2/000019.mp4,test/val_plains/2/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +93,93,127,test/val_desert/3/000049.mp4,test/val_desert/3/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +94,94,370,test/val_iceplains/2/000048.mp4,test/val_iceplains/2/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +95,95,293,test/val_iceplains/2/000018.mp4,test/val_iceplains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +96,96,251,test/val_iceplains/1/000003.mp4,test/val_iceplains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +97,97,441,test/val_plains/1/000031.mp4,test/val_plains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +98,98,69,test/val_desert/2/000026.mp4,test/val_desert/2/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +99,99,286,test/val_iceplains/2/000015.mp4,test/val_iceplains/2/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +100,100,438,test/val_plains/0/000029.mp4,test/val_plains/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +101,101,376,test/val_plains/0/000002.mp4,test/val_plains/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +102,102,163,test/val_savanna/2/000014.mp4,test/val_savanna/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +103,103,31,test/val_desert/1/000013.mp4,test/val_desert/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +104,104,241,test/val_savanna/0/000049.mp4,test/val_savanna/0/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +105,105,458,test/val_plains/0/000040.mp4,test/val_plains/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +106,106,371,test/val_iceplains/1/000049.mp4,test/val_iceplains/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +107,107,230,test/val_savanna/3/000045.mp4,test/val_savanna/3/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +108,108,372,test/val_iceplains/2/000049.mp4,test/val_iceplains/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +109,109,223,test/val_savanna/1/000042.mp4,test/val_savanna/1/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +110,110,68,test/val_desert/0/000026.mp4,test/val_desert/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +111,111,461,test/val_plains/2/000042.mp4,test/val_plains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +112,112,151,test/val_savanna/1/000009.mp4,test/val_savanna/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +113,113,72,test/val_desert/3/000027.mp4,test/val_desert/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +114,114,427,test/val_plains/0/000024.mp4,test/val_plains/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +115,115,317,test/val_iceplains/3/000029.mp4,test/val_iceplains/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +116,116,249,test/val_iceplains/2/000001.mp4,test/val_iceplains/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +117,117,364,test/val_iceplains/2/000046.mp4,test/val_iceplains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +118,118,96,test/val_desert/2/000037.mp4,test/val_desert/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +119,119,190,test/val_savanna/3/000029.mp4,test/val_savanna/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +120,120,227,test/val_savanna/2/000044.mp4,test/val_savanna/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +121,121,244,test/val_iceplains/0/000000.mp4,test/val_iceplains/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +122,122,38,test/val_desert/3/000015.mp4,test/val_desert/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +123,123,291,test/val_iceplains/2/000017.mp4,test/val_iceplains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +124,124,85,test/val_desert/2/000032.mp4,test/val_desert/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +125,125,358,test/val_iceplains/0/000044.mp4,test/val_iceplains/0/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +126,126,50,test/val_desert/1/000020.mp4,test/val_desert/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +127,127,207,test/val_savanna/0/000036.mp4,test/val_savanna/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +128,128,330,test/val_iceplains/2/000034.mp4,test/val_iceplains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +129,129,132,test/val_savanna/0/000001.mp4,test/val_savanna/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +130,130,423,test/val_plains/3/000021.mp4,test/val_plains/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +131,131,66,test/val_desert/2/000025.mp4,test/val_desert/2/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +132,132,153,test/val_savanna/3/000009.mp4,test/val_savanna/3/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +133,133,382,test/val_plains/0/000004.mp4,test/val_plains/0/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +134,134,143,test/val_savanna/2/000005.mp4,test/val_savanna/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +135,135,348,test/val_iceplains/1/000040.mp4,test/val_iceplains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +136,136,384,test/val_plains/2/000005.mp4,test/val_plains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +137,137,30,test/val_desert/0/000012.mp4,test/val_desert/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +138,138,203,test/val_savanna/2/000034.mp4,test/val_savanna/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +139,139,179,test/val_savanna/0/000022.mp4,test/val_savanna/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +140,140,253,test/val_iceplains/2/000003.mp4,test/val_iceplains/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +141,141,289,test/val_iceplains/2/000016.mp4,test/val_iceplains/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +142,142,422,test/val_plains/2/000020.mp4,test/val_plains/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +143,143,20,test/val_desert/0/000008.mp4,test/val_desert/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +144,144,130,test/val_savanna/0/000000.mp4,test/val_savanna/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +145,145,296,test/val_iceplains/1/000020.mp4,test/val_iceplains/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +146,146,278,test/val_iceplains/3/000013.mp4,test/val_iceplains/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +147,147,269,test/val_iceplains/0/000009.mp4,test/val_iceplains/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +148,148,314,test/val_iceplains/0/000028.mp4,test/val_iceplains/0/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +149,149,281,test/val_iceplains/0/000014.mp4,test/val_iceplains/0/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +150,150,149,test/val_savanna/0/000008.mp4,test/val_savanna/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +151,151,184,test/val_savanna/0/000024.mp4,test/val_savanna/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +152,152,407,test/val_plains/1/000014.mp4,test/val_plains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +153,153,304,test/val_iceplains/2/000023.mp4,test/val_iceplains/2/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +154,154,208,test/val_savanna/3/000037.mp4,test/val_savanna/3/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +155,155,182,test/val_savanna/3/000023.mp4,test/val_savanna/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +156,156,394,test/val_plains/1/000009.mp4,test/val_plains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +157,157,14,test/val_desert/0/000006.mp4,test/val_desert/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +158,158,55,test/val_desert/3/000021.mp4,test/val_desert/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +159,159,75,test/val_desert/2/000028.mp4,test/val_desert/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +160,160,319,test/val_iceplains/1/000031.mp4,test/val_iceplains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +161,161,39,test/val_desert/1/000016.mp4,test/val_desert/1/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +162,162,418,test/val_plains/1/000019.mp4,test/val_plains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +163,163,389,test/val_plains/3/000007.mp4,test/val_plains/3/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +164,164,409,test/val_plains/2/000014.mp4,test/val_plains/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +165,165,344,test/val_iceplains/2/000038.mp4,test/val_iceplains/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +166,166,396,test/val_plains/2/000009.mp4,test/val_plains/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +167,167,67,test/val_desert/1/000026.mp4,test/val_desert/1/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +168,168,99,test/val_desert/3/000038.mp4,test/val_desert/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +169,169,288,test/val_iceplains/3/000016.mp4,test/val_iceplains/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +170,170,97,test/val_desert/1/000038.mp4,test/val_desert/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +171,171,178,test/val_savanna/2/000021.mp4,test/val_savanna/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +172,172,264,test/val_iceplains/1/000007.mp4,test/val_iceplains/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +173,173,266,test/val_iceplains/0/000008.mp4,test/val_iceplains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +174,174,303,test/val_iceplains/1/000023.mp4,test/val_iceplains/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +175,175,442,test/val_plains/0/000031.mp4,test/val_plains/0/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +176,176,233,test/val_savanna/1/000047.mp4,test/val_savanna/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +177,177,240,test/val_savanna/1/000049.mp4,test/val_savanna/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +178,178,42,test/val_desert/0/000017.mp4,test/val_desert/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +179,179,369,test/val_iceplains/3/000048.mp4,test/val_iceplains/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +180,180,398,test/val_plains/3/000010.mp4,test/val_plains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +181,181,435,test/val_plains/2/000027.mp4,test/val_plains/2/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +182,182,242,test/val_savanna/2/000049.mp4,test/val_savanna/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +183,183,361,test/val_iceplains/2/000045.mp4,test/val_iceplains/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +184,184,470,test/val_plains/2/000047.mp4,test/val_plains/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +185,185,356,test/val_iceplains/2/000042.mp4,test/val_iceplains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +186,186,468,test/val_plains/2/000046.mp4,test/val_plains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +187,187,113,test/val_desert/0/000043.mp4,test/val_desert/0/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +188,188,19,test/val_desert/1/000008.mp4,test/val_desert/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +189,189,199,test/val_savanna/0/000033.mp4,test/val_savanna/0/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +190,190,118,test/val_desert/2/000045.mp4,test/val_desert/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +191,191,433,test/val_plains/0/000027.mp4,test/val_plains/0/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +192,192,328,test/val_iceplains/0/000034.mp4,test/val_iceplains/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +193,193,213,test/val_savanna/0/000039.mp4,test/val_savanna/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +194,194,339,test/val_iceplains/0/000037.mp4,test/val_iceplains/0/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +195,195,466,test/val_plains/1/000046.mp4,test/val_plains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +196,196,355,test/val_iceplains/3/000042.mp4,test/val_iceplains/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +197,197,336,test/val_iceplains/3/000036.mp4,test/val_iceplains/3/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +198,198,420,test/val_plains/0/000020.mp4,test/val_plains/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +199,199,410,test/val_plains/1/000015.mp4,test/val_plains/1/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +200,200,27,test/val_desert/1/000011.mp4,test/val_desert/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +201,201,86,test/val_desert/1/000033.mp4,test/val_desert/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +202,202,210,test/val_savanna/3/000038.mp4,test/val_savanna/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +203,203,103,test/val_desert/3/000039.mp4,test/val_desert/3/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +204,204,245,test/val_iceplains/3/000000.mp4,test/val_iceplains/3/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +205,205,25,test/val_desert/2/000009.mp4,test/val_desert/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +206,206,340,test/val_iceplains/2/000037.mp4,test/val_iceplains/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +207,207,167,test/val_savanna/0/000016.mp4,test/val_savanna/0/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +208,208,424,test/val_plains/0/000022.mp4,test/val_plains/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +209,209,448,test/val_plains/2/000033.mp4,test/val_plains/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +210,210,386,test/val_plains/0/000006.mp4,test/val_plains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +211,211,392,test/val_plains/3/000008.mp4,test/val_plains/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +212,212,105,test/val_desert/1/000040.mp4,test/val_desert/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +213,213,267,test/val_iceplains/2/000008.mp4,test/val_iceplains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +214,214,90,test/val_desert/0/000034.mp4,test/val_desert/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +215,215,426,test/val_plains/0/000023.mp4,test/val_plains/0/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +216,216,168,test/val_savanna/3/000016.mp4,test/val_savanna/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +217,217,218,test/val_savanna/3/000040.mp4,test/val_savanna/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +218,218,445,test/val_plains/0/000032.mp4,test/val_plains/0/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +219,219,48,test/val_desert/3/000019.mp4,test/val_desert/3/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +220,220,104,test/val_desert/2/000039.mp4,test/val_desert/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +221,221,206,test/val_savanna/2/000035.mp4,test/val_savanna/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +222,222,228,test/val_savanna/1/000045.mp4,test/val_savanna/1/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +223,223,200,test/val_savanna/2/000033.mp4,test/val_savanna/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +224,224,62,test/val_desert/2/000024.mp4,test/val_desert/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +225,225,121,test/val_desert/1/000047.mp4,test/val_desert/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +226,226,147,test/val_savanna/0/000007.mp4,test/val_savanna/0/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +227,227,146,test/val_savanna/2/000006.mp4,test/val_savanna/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +228,228,225,test/val_savanna/1/000043.mp4,test/val_savanna/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +229,229,313,test/val_iceplains/3/000027.mp4,test/val_iceplains/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +230,230,354,test/val_iceplains/0/000042.mp4,test/val_iceplains/0/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +231,231,367,test/val_iceplains/1/000048.mp4,test/val_iceplains/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +232,232,297,test/val_iceplains/3/000020.mp4,test/val_iceplains/3/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +233,233,124,test/val_desert/1/000048.mp4,test/val_desert/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +234,234,171,test/val_savanna/2/000017.mp4,test/val_savanna/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +235,235,342,test/val_iceplains/0/000038.mp4,test/val_iceplains/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +236,236,60,test/val_desert/3/000023.mp4,test/val_desert/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +237,237,131,test/val_savanna/1/000001.mp4,test/val_savanna/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +238,238,365,test/val_iceplains/1/000047.mp4,test/val_iceplains/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +239,239,315,test/val_iceplains/2/000028.mp4,test/val_iceplains/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +240,240,220,test/val_savanna/1/000041.mp4,test/val_savanna/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +241,241,8,test/val_desert/1/000003.mp4,test/val_desert/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +242,242,261,test/val_iceplains/0/000006.mp4,test/val_iceplains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +243,243,17,test/val_desert/1/000007.mp4,test/val_desert/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +244,244,462,test/val_plains/1/000043.mp4,test/val_plains/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +245,245,454,test/val_plains/3/000038.mp4,test/val_plains/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +246,246,351,test/val_iceplains/1/000041.mp4,test/val_iceplains/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +247,247,383,test/val_plains/3/000005.mp4,test/val_plains/3/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +248,248,18,test/val_desert/2/000007.mp4,test/val_desert/2/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +249,249,457,test/val_plains/1/000040.mp4,test/val_plains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +250,250,326,test/val_iceplains/3/000033.mp4,test/val_iceplains/3/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +251,251,254,test/val_iceplains/1/000004.mp4,test/val_iceplains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +252,252,375,test/val_plains/1/000002.mp4,test/val_plains/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +253,253,141,test/val_savanna/0/000005.mp4,test/val_savanna/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +254,254,306,test/val_iceplains/2/000024.mp4,test/val_iceplains/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +255,255,10,test/val_desert/2/000003.mp4,test/val_desert/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +256,256,173,test/val_savanna/0/000020.mp4,test/val_savanna/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +257,257,320,test/val_iceplains/1/000032.mp4,test/val_iceplains/1/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +258,258,115,test/val_desert/1/000044.mp4,test/val_desert/1/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +259,259,341,test/val_iceplains/1/000038.mp4,test/val_iceplains/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +260,260,61,test/val_desert/3/000024.mp4,test/val_desert/3/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +261,261,231,test/val_savanna/1/000046.mp4,test/val_savanna/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +262,262,224,test/val_savanna/3/000042.mp4,test/val_savanna/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +263,263,292,test/val_iceplains/3/000018.mp4,test/val_iceplains/3/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +264,264,334,test/val_iceplains/1/000036.mp4,test/val_iceplains/1/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +265,265,159,test/val_savanna/0/000013.mp4,test/val_savanna/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +266,266,102,test/val_desert/0/000039.mp4,test/val_desert/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +267,267,362,test/val_iceplains/1/000046.mp4,test/val_iceplains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +268,268,322,test/val_iceplains/3/000032.mp4,test/val_iceplains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +269,269,381,test/val_plains/1/000004.mp4,test/val_plains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +270,270,84,test/val_desert/3/000032.mp4,test/val_desert/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +271,271,243,test/val_iceplains/1/000000.mp4,test/val_iceplains/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +272,272,65,test/val_desert/3/000025.mp4,test/val_desert/3/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +273,273,64,test/val_desert/0/000025.mp4,test/val_desert/0/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +274,274,272,test/val_iceplains/2/000010.mp4,test/val_iceplains/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +275,275,399,test/val_plains/0/000011.mp4,test/val_plains/0/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +276,276,248,test/val_iceplains/3/000001.mp4,test/val_iceplains/3/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +277,277,43,test/val_desert/3/000017.mp4,test/val_desert/3/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +278,278,393,test/val_plains/2/000008.mp4,test/val_plains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +279,279,123,test/val_desert/2/000047.mp4,test/val_desert/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +280,280,29,test/val_desert/2/000011.mp4,test/val_desert/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +281,281,122,test/val_desert/0/000047.mp4,test/val_desert/0/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +282,282,157,test/val_savanna/3/000011.mp4,test/val_savanna/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +283,283,333,test/val_iceplains/2/000035.mp4,test/val_iceplains/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +284,284,307,test/val_iceplains/1/000025.mp4,test/val_iceplains/1/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +285,285,9,test/val_desert/0/000003.mp4,test/val_desert/0/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +286,286,250,test/val_iceplains/2/000002.mp4,test/val_iceplains/2/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +287,287,165,test/val_savanna/3/000015.mp4,test/val_savanna/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +288,288,0,test/val_desert/1/000000.mp4,test/val_desert/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +289,289,271,test/val_iceplains/3/000010.mp4,test/val_iceplains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +290,290,276,test/val_iceplains/1/000013.mp4,test/val_iceplains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +291,291,82,test/val_desert/1/000031.mp4,test/val_desert/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +292,292,191,test/val_savanna/2/000029.mp4,test/val_savanna/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +293,293,374,test/val_plains/1/000001.mp4,test/val_plains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +294,294,446,test/val_plains/3/000032.mp4,test/val_plains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +295,295,378,test/val_plains/1/000003.mp4,test/val_plains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +296,296,152,test/val_savanna/0/000009.mp4,test/val_savanna/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +297,297,54,test/val_desert/0/000021.mp4,test/val_desert/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +298,298,259,test/val_iceplains/2/000005.mp4,test/val_iceplains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +299,299,91,test/val_desert/2/000034.mp4,test/val_desert/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.sha256 b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..065350069e5b0f46e33cd51ad63f60aa956822fd --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_age_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.sha256 @@ -0,0 +1 @@ +ef4e4de9d3975d8a7e724bb609e57f20186cd1f7285457c00fb85c1aca973b99 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/checkpoint_sha256.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/checkpoint_sha256.csv new file mode 100644 index 0000000000000000000000000000000000000000..0cbabb25f31be55688d79358c5cfed3fa5754432 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/checkpoint_sha256.csv @@ -0,0 +1,2 @@ +path,size_bytes,sha256 +/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,11853520525,c33c96daf31163d9efcdcf800479ad6b599353fdd81929b1ebd10af9044e75ec diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/evaluation.patch b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/evaluation.patch new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..3cfd08bc49f4159d2bd5cace6409ac92956d78e1 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.csv @@ -0,0 +1,51 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.sha256 b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..67b0dc2042229c0768d324282215bdf64bebe28f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.sha256 @@ -0,0 +1 @@ +f9536911e434610cce23306ff99575db53cd7fc29e3df841da4b6631bf418634 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/git_provenance.json b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/git_provenance.json new file mode 100644 index 0000000000000000000000000000000000000000..fabe691fc08806f957f2445d4a52b2f38d7f07d8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/git_provenance.json @@ -0,0 +1,5 @@ +{ + "branch": "dememwm-unified-sourcefree-age-balanced", + "head": "732555281e86c4fdbb71c98a000d648c3edc3109", + "status": "?? results/\n?? train_dememwm_full_berzelius_gate_off_h200.sh\n?? train_dememwm_separate_mem_gatefix_4layers512_valid500_berzelius_h200.sh" +} diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/launcher.sh b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/launcher.sh new file mode 100644 index 0000000000000000000000000000000000000000..db637f1236690823d28fd0f38aa79441ca248693 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/launcher.sh @@ -0,0 +1,182 @@ +#!/usr/bin/env bash +#SBATCH --job-name=demem-sf-countoff-lh300 +#SBATCH --nodes=1 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=14 +#SBATCH --gres=gpu:8 +#SBATCH --time=48:00:00 +#SBATCH --account=berzelius-2026-196 +#SBATCH --array=42-43%1 +#SBATCH --chdir=/proj/cvl/users/x_fahkh2/DeMemWM +#SBATCH --output=/proj/cvl/users/x_fahkh2/caches/slurm-%x-%A_%a.out +#SBATCH --error=/proj/cvl/users/x_fahkh2/caches/slurm-%x-%A_%a.out + +set -eo pipefail + +module load buildenv-gcccuda/12.9.1-gcc11 +source $(conda info --base)/etc/profile.d/conda.sh + +REPO=/proj/cvl/users/x_fahkh2/DeMemWM +DATA_DIR=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +FEATURE_DIR=$DATA_DIR/vae_features +CACHE_DIR=/proj/cvl/users/x_fahkh2/caches +TRAIN_RUN=train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k +MODEL=dememwm_sourcefree_countcorr_off_step200k +CHECKPOINT=$REPO/results/$TRAIN_RUN/checkpoints/epoch3_step200000.ckpt +OUTPUT_ROOT=$REPO/outputs/dememwm_sourcefree_long_horizon +MANIFEST=$REPO/evaluation/manifests/minecraft_test_300.csv +FID_MANIFEST=$REPO/evaluation/manifests/minecraft_test_fid50.csv +FID_BATCH_SIZE=32 + +export PYTHONPATH="$REPO:${PYTHONPATH:-}" +export HF_HOME=$CACHE_DIR +export TORCH_HOME=$CACHE_DIR +export PIP_CACHE_DIR=$CACHE_DIR +export TMPDIR=/scratch/local/$USER/dememwm_eval_$SLURM_JOB_ID +export TRITON_CACHE_DIR=$TMPDIR +export CUDA_HOME=$CUDA_ROOT +export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +export WANDB_DISABLED=true +export HYDRA_FULL_ERROR=1 +export PYTHONWARNINGS=ignore +export PYTHONUNBUFFERED=1 +export OMP_NUM_THREADS=1 +export MKL_NUM_THREADS=1 + +cd "$REPO" +mkdir -p "$TMPDIR" "$OUTPUT_ROOT" + +MASTER_SEED=$SLURM_ARRAY_TASK_ID +RUN_NAME=${MODEL}_minecraft_test300_ctx600_roll500_bs1_seed${MASTER_SEED} +OUTPUT_DIR=$OUTPUT_ROOT/$RUN_NAME +WORKERS_DIR=$OUTPUT_DIR/workers +RANGES=( + 0:10 10:20 20:30 30:40 + 40:50 50:60 60:70 70:80 + 80:90 90:100 100:110 110:120 + 120:129 129:138 138:147 147:156 + 156:165 165:174 174:183 183:192 + 192:201 201:210 210:219 219:228 + 228:237 237:246 246:255 255:264 + 264:273 273:282 282:291 291:300 +) + +python long_horizon_eval.py verify-manifest --manifest "$MANIFEST" --clips 300 +python long_horizon_eval.py verify-manifest --manifest "$FID_MANIFEST" --clips 50 +python long_horizon_eval.py init-run \ + --output "$OUTPUT_DIR" \ + --repo "$REPO" \ + --manifest "$MANIFEST" \ + --fid-manifest "$FID_MANIFEST" \ + --model "$MODEL" \ + --master-seed "$MASTER_SEED" \ + --batch-size 1 \ + --checkpoint "$CHECKPOINT" \ + "${RANGES[@]/#/--worker-range=}" + +mkdir -p "$WORKERS_DIR" +cp evaluate_dememwm_sourcefree_countcorr_off_long_horizon_8gpu.sh "$OUTPUT_DIR/launcher.sh" + +PIDS=() +for WORKER_ID in "${!RANGES[@]}"; do + GPU_ID=$((WORKER_ID / 4)) + START=${RANGES[$WORKER_ID]%:*} + END=${RANGES[$WORKER_ID]#*:} + COUNT=$((END - START)) + WORKER_DIR=$(printf '%s/worker_%02d' "$WORKERS_DIR" "$WORKER_ID") + mkdir -p "$WORKER_DIR" + + CUDA_VISIBLE_DEVICES=$GPU_ID python -m main \ + +name="$RUN_NAME" \ + hydra.run.dir="$WORKER_DIR/hydra/run" \ + load="$CHECKPOINT" \ + auto_resume=false \ + +customized_load=false \ + wandb.mode=disabled \ + "experiment.tasks=[test]" \ + experiment.num_nodes=1 \ + +seed="$MASTER_SEED" \ + dataset=video_minecraft_dememwm_latent \ + dataset.save_dir="$DATA_DIR" \ + dataset.precomputed_feature_dir="$FEATURE_DIR" \ + dataset.n_frames=8 \ + dataset.n_frames_valid=1100 \ + dataset.context_length=100 \ + dataset.single_eval_clip=true \ + dataset.memory_selection.enabled=true \ + dataset.memory_selection.causal=true \ + dataset.memory_selection.max_anchor_frames=2 \ + dataset.memory_selection.max_dynamic_frames=4 \ + dataset.memory_selection.max_revisit_frames=2 \ + dataset.memory_selection.training_dropout=0.1 \ + dataset.memory_selection.dynamic.selection_policy=multiview \ + dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse \ + +dataset.seed=42 \ + +dataset.wo_updown=false \ + algorithm=dememwm_base \ + algorithm.causal=true \ + algorithm.n_tokens=8 \ + algorithm.context_frames=600 \ + algorithm.metric_report_segment=100 \ + algorithm.use_memory_attention=true \ + algorithm.anchor_zero_geometry=false \ + algorithm.memory_role_embedding=true \ + algorithm.memory_source_embedding=false \ + algorithm.memory_attention_key_only_geometry=true \ + algorithm.memory_age.enabled=true \ + algorithm.memory_age.num_buckets=32 \ + algorithm.memory_age.num_exact_buckets=17 \ + algorithm.memory_age.overflow_start=100 \ + algorithm.memory_stream_count_correction=false \ + algorithm.noise_route.anchor=high \ + algorithm.noise_route.dynamic=low \ + algorithm.noise_route.revisit=all \ + algorithm.memory_source_simulation.training_generated_suffix_max_frames=null \ + algorithm.memory_noise.enabled=true \ + algorithm.memory_noise.mode=source_free_logsnr \ + algorithm.memory_noise.stable_probability=0.25 \ + algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 \ + algorithm.memory_noise.validation_noisy_memory=false \ + algorithm.chunk_size=1 \ + algorithm.diffusion.sampling_timesteps=20 \ + algorithm.diffusion.ddim_sampling_eta=0.0 \ + algorithm.log_video=false \ + algorithm.save_local=true \ + +algorithm.local_save_dir="$WORKER_DIR" \ + +algorithm.lpips_batch_size=16 \ + +algorithm.eval_manifest_path="$MANIFEST" \ + +algorithm.eval_manifest_start="$START" \ + +algorithm.eval_manifest_count="$COUNT" \ + +algorithm.eval_master_seed="$MASTER_SEED" \ + +algorithm.eval_batch_size=1 \ + +experiment.test.manifest_path="$MANIFEST" \ + +experiment.test.manifest_start="$START" \ + +experiment.test.manifest_count="$COUNT" \ + experiment.test.batch_size=1 \ + experiment.test.limit_batch="$COUNT" \ + experiment.test.data.num_workers=2 \ + experiment.test.inference_mode=false \ + > "$OUTPUT_DIR/worker_${WORKER_ID}.log" 2>&1 & + PIDS+=("$!") +done + +FAILED=0 +for PID in "${PIDS[@]}"; do + wait "$PID" || FAILED=1 +done +test "$FAILED" -eq 0 + +python -c 'from omegaconf import OmegaConf; import sys; cfg=OmegaConf.load(sys.argv[1]); OmegaConf.save(cfg, sys.argv[2], resolve=True)' \ + "$WORKERS_DIR/worker_00/hydra/run/.hydra/config.yaml" "$OUTPUT_DIR/resolved_config.yaml" +python long_horizon_eval.py aggregate \ + --manifest "$MANIFEST" \ + --workers-root "$WORKERS_DIR" \ + --output "$OUTPUT_DIR" \ + --master-seed "$MASTER_SEED" \ + --workers 32 +srun --exclusive --exact --nodes=1 --ntasks=1 --gpus-per-task=1 --cpus-per-task=8 \ + python calculate_fid.py \ + --run-dir "$OUTPUT_DIR" \ + --cache-dir "$CACHE_DIR/fid_real_stats" \ + --batch-size "$FID_BATCH_SIZE" \ + --device cuda diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/protocol.json b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/protocol.json new file mode 100644 index 0000000000000000000000000000000000000000..14094859dd0d1f4667ee1d934550f6208126a0b1 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/protocol.json @@ -0,0 +1,122 @@ +{ + "model": "dememwm_sourcefree_countcorr_off_step200k", + "master_seed": 42, + "sample_seed_formula": "master_seed * 1000000 + global_test_index", + "test_clips": 300, + "source_offset": 100, + "context_range": [ + 100, + 700 + ], + "target_range": [ + 700, + 1200 + ], + "context_frames": 600, + "generated_frames": 500, + "local_window": 8, + "ddim_steps": 20, + "ddim_eta": 0.0, + "per_device_batch_size": 1, + "worker_ranges": [ + "0:10", + "10:20", + "20:30", + "30:40", + "40:50", + "50:60", + "60:70", + "70:80", + "80:90", + "90:100", + "100:110", + "110:120", + "120:129", + "129:138", + "138:147", + "147:156", + "156:165", + "165:174", + "174:183", + "183:192", + "192:201", + "201:210", + "210:219", + "219:228", + "228:237", + "237:246", + "246:255", + "255:264", + "264:273", + "273:282", + "282:291", + "291:300" + ], + "distribution": "32 disjoint standalone batch-1 workers; four per GPU; no DistributedSampler", + "generated_frame_bins": [ + [ + 0, + 100 + ], + [ + 100, + 200 + ], + [ + 200, + 300 + ], + [ + 300, + 400 + ], + [ + 400, + 500 + ] + ], + "reconstruction_metrics": { + "names": [ + "mse", + "psnr", + "lpips" + ], + "population_per_bin": { + "clips": 300, + "frames": 30000 + }, + "population_overall": { + "clips": 300, + "frames": 150000 + }, + "lpips_input_range": [ + 0.0, + 1.0 + ], + "lpips_normalize": true + }, + "frame_fid": { + "implementation": "TorchMetrics FrechetInceptionDistance", + "feature": 2048, + "normalize": false, + "input": "uint8 RGB [0,255]", + "official_compatible_bins": { + "clips": 50, + "frames_per_bin": 5000 + }, + "supplementary_pooled": { + "clips": 50, + "frames": 25000 + }, + "full_300_bins": { + "clips": 300, + "frames_per_bin": 30000 + }, + "full_300_pooled": { + "clips": 300, + "frames": 150000 + } + }, + "manifest_sha256": "ef4e4de9d3975d8a7e724bb609e57f20186cd1f7285457c00fb85c1aca973b99", + "fid50_manifest_sha256": "f9536911e434610cce23306ff99575db53cd7fc29e3df841da4b6631bf418634" +} diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..dae70b4071c2d2a834268c2a26f061d308f11f2c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.csv @@ -0,0 +1,301 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +50,50,212,test/val_savanna/1/000039.mp4,test/val_savanna/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +51,51,4,test/val_desert/0/000001.mp4,test/val_desert/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +52,52,7,test/val_desert/0/000002.mp4,test/val_desert/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +53,53,310,test/val_iceplains/0/000026.mp4,test/val_iceplains/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +54,54,284,test/val_iceplains/0/000015.mp4,test/val_iceplains/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +55,55,109,test/val_desert/3/000041.mp4,test/val_desert/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +56,56,45,test/val_desert/0/000018.mp4,test/val_desert/0/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +57,57,93,test/val_desert/0/000035.mp4,test/val_desert/0/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +58,58,400,test/val_plains/3/000011.mp4,test/val_plains/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +59,59,300,test/val_iceplains/2/000021.mp4,test/val_iceplains/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +60,60,226,test/val_savanna/2/000043.mp4,test/val_savanna/2/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +61,61,417,test/val_plains/2/000018.mp4,test/val_plains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +62,62,120,test/val_desert/3/000046.mp4,test/val_desert/3/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +63,63,106,test/val_desert/3/000040.mp4,test/val_desert/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +64,64,345,test/val_iceplains/1/000039.mp4,test/val_iceplains/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +65,65,169,test/val_savanna/2/000016.mp4,test/val_savanna/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +66,66,312,test/val_iceplains/1/000027.mp4,test/val_iceplains/1/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +67,67,368,test/val_iceplains/0/000048.mp4,test/val_iceplains/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +68,68,449,test/val_plains/1/000034.mp4,test/val_plains/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +69,69,11,test/val_desert/1/000004.mp4,test/val_desert/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +70,70,6,test/val_desert/2/000001.mp4,test/val_desert/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +71,71,257,test/val_iceplains/0/000005.mp4,test/val_iceplains/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +72,72,262,test/val_iceplains/3/000006.mp4,test/val_iceplains/3/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +73,73,164,test/val_savanna/0/000015.mp4,test/val_savanna/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +74,74,337,test/val_iceplains/2/000036.mp4,test/val_iceplains/2/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +75,75,95,test/val_desert/0/000036.mp4,test/val_desert/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +76,76,277,test/val_iceplains/0/000013.mp4,test/val_iceplains/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +77,77,26,test/val_desert/1/000010.mp4,test/val_desert/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +78,78,77,test/val_desert/0/000029.mp4,test/val_desert/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +79,79,294,test/val_iceplains/1/000019.mp4,test/val_iceplains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +80,80,34,test/val_desert/1/000014.mp4,test/val_desert/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +81,81,188,test/val_savanna/1/000029.mp4,test/val_savanna/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +82,82,170,test/val_savanna/0/000017.mp4,test/val_savanna/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +83,83,415,test/val_plains/2/000017.mp4,test/val_plains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +84,84,145,test/val_savanna/0/000006.mp4,test/val_savanna/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +85,85,133,test/val_savanna/1/000002.mp4,test/val_savanna/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +86,86,347,test/val_iceplains/2/000039.mp4,test/val_iceplains/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +87,87,209,test/val_savanna/0/000038.mp4,test/val_savanna/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +88,88,404,test/val_plains/1/000013.mp4,test/val_plains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +89,89,211,test/val_savanna/2/000038.mp4,test/val_savanna/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +90,90,464,test/val_plains/2/000044.mp4,test/val_plains/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +91,91,162,test/val_savanna/1/000014.mp4,test/val_savanna/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +92,92,419,test/val_plains/2/000019.mp4,test/val_plains/2/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +93,93,127,test/val_desert/3/000049.mp4,test/val_desert/3/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +94,94,370,test/val_iceplains/2/000048.mp4,test/val_iceplains/2/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +95,95,293,test/val_iceplains/2/000018.mp4,test/val_iceplains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +96,96,251,test/val_iceplains/1/000003.mp4,test/val_iceplains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +97,97,441,test/val_plains/1/000031.mp4,test/val_plains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +98,98,69,test/val_desert/2/000026.mp4,test/val_desert/2/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +99,99,286,test/val_iceplains/2/000015.mp4,test/val_iceplains/2/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +100,100,438,test/val_plains/0/000029.mp4,test/val_plains/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +101,101,376,test/val_plains/0/000002.mp4,test/val_plains/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +102,102,163,test/val_savanna/2/000014.mp4,test/val_savanna/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +103,103,31,test/val_desert/1/000013.mp4,test/val_desert/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +104,104,241,test/val_savanna/0/000049.mp4,test/val_savanna/0/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +105,105,458,test/val_plains/0/000040.mp4,test/val_plains/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +106,106,371,test/val_iceplains/1/000049.mp4,test/val_iceplains/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +107,107,230,test/val_savanna/3/000045.mp4,test/val_savanna/3/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +108,108,372,test/val_iceplains/2/000049.mp4,test/val_iceplains/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +109,109,223,test/val_savanna/1/000042.mp4,test/val_savanna/1/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +110,110,68,test/val_desert/0/000026.mp4,test/val_desert/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +111,111,461,test/val_plains/2/000042.mp4,test/val_plains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +112,112,151,test/val_savanna/1/000009.mp4,test/val_savanna/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +113,113,72,test/val_desert/3/000027.mp4,test/val_desert/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +114,114,427,test/val_plains/0/000024.mp4,test/val_plains/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +115,115,317,test/val_iceplains/3/000029.mp4,test/val_iceplains/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +116,116,249,test/val_iceplains/2/000001.mp4,test/val_iceplains/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +117,117,364,test/val_iceplains/2/000046.mp4,test/val_iceplains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +118,118,96,test/val_desert/2/000037.mp4,test/val_desert/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +119,119,190,test/val_savanna/3/000029.mp4,test/val_savanna/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +120,120,227,test/val_savanna/2/000044.mp4,test/val_savanna/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +121,121,244,test/val_iceplains/0/000000.mp4,test/val_iceplains/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +122,122,38,test/val_desert/3/000015.mp4,test/val_desert/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +123,123,291,test/val_iceplains/2/000017.mp4,test/val_iceplains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +124,124,85,test/val_desert/2/000032.mp4,test/val_desert/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +125,125,358,test/val_iceplains/0/000044.mp4,test/val_iceplains/0/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +126,126,50,test/val_desert/1/000020.mp4,test/val_desert/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +127,127,207,test/val_savanna/0/000036.mp4,test/val_savanna/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +128,128,330,test/val_iceplains/2/000034.mp4,test/val_iceplains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +129,129,132,test/val_savanna/0/000001.mp4,test/val_savanna/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +130,130,423,test/val_plains/3/000021.mp4,test/val_plains/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +131,131,66,test/val_desert/2/000025.mp4,test/val_desert/2/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +132,132,153,test/val_savanna/3/000009.mp4,test/val_savanna/3/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +133,133,382,test/val_plains/0/000004.mp4,test/val_plains/0/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +134,134,143,test/val_savanna/2/000005.mp4,test/val_savanna/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +135,135,348,test/val_iceplains/1/000040.mp4,test/val_iceplains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +136,136,384,test/val_plains/2/000005.mp4,test/val_plains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +137,137,30,test/val_desert/0/000012.mp4,test/val_desert/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +138,138,203,test/val_savanna/2/000034.mp4,test/val_savanna/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +139,139,179,test/val_savanna/0/000022.mp4,test/val_savanna/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +140,140,253,test/val_iceplains/2/000003.mp4,test/val_iceplains/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +141,141,289,test/val_iceplains/2/000016.mp4,test/val_iceplains/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +142,142,422,test/val_plains/2/000020.mp4,test/val_plains/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +143,143,20,test/val_desert/0/000008.mp4,test/val_desert/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +144,144,130,test/val_savanna/0/000000.mp4,test/val_savanna/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +145,145,296,test/val_iceplains/1/000020.mp4,test/val_iceplains/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +146,146,278,test/val_iceplains/3/000013.mp4,test/val_iceplains/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +147,147,269,test/val_iceplains/0/000009.mp4,test/val_iceplains/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +148,148,314,test/val_iceplains/0/000028.mp4,test/val_iceplains/0/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +149,149,281,test/val_iceplains/0/000014.mp4,test/val_iceplains/0/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +150,150,149,test/val_savanna/0/000008.mp4,test/val_savanna/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +151,151,184,test/val_savanna/0/000024.mp4,test/val_savanna/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +152,152,407,test/val_plains/1/000014.mp4,test/val_plains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +153,153,304,test/val_iceplains/2/000023.mp4,test/val_iceplains/2/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +154,154,208,test/val_savanna/3/000037.mp4,test/val_savanna/3/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +155,155,182,test/val_savanna/3/000023.mp4,test/val_savanna/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +156,156,394,test/val_plains/1/000009.mp4,test/val_plains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +157,157,14,test/val_desert/0/000006.mp4,test/val_desert/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +158,158,55,test/val_desert/3/000021.mp4,test/val_desert/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +159,159,75,test/val_desert/2/000028.mp4,test/val_desert/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +160,160,319,test/val_iceplains/1/000031.mp4,test/val_iceplains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +161,161,39,test/val_desert/1/000016.mp4,test/val_desert/1/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +162,162,418,test/val_plains/1/000019.mp4,test/val_plains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +163,163,389,test/val_plains/3/000007.mp4,test/val_plains/3/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +164,164,409,test/val_plains/2/000014.mp4,test/val_plains/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +165,165,344,test/val_iceplains/2/000038.mp4,test/val_iceplains/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +166,166,396,test/val_plains/2/000009.mp4,test/val_plains/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +167,167,67,test/val_desert/1/000026.mp4,test/val_desert/1/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +168,168,99,test/val_desert/3/000038.mp4,test/val_desert/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +169,169,288,test/val_iceplains/3/000016.mp4,test/val_iceplains/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +170,170,97,test/val_desert/1/000038.mp4,test/val_desert/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +171,171,178,test/val_savanna/2/000021.mp4,test/val_savanna/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +172,172,264,test/val_iceplains/1/000007.mp4,test/val_iceplains/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +173,173,266,test/val_iceplains/0/000008.mp4,test/val_iceplains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +174,174,303,test/val_iceplains/1/000023.mp4,test/val_iceplains/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +175,175,442,test/val_plains/0/000031.mp4,test/val_plains/0/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +176,176,233,test/val_savanna/1/000047.mp4,test/val_savanna/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +177,177,240,test/val_savanna/1/000049.mp4,test/val_savanna/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +178,178,42,test/val_desert/0/000017.mp4,test/val_desert/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +179,179,369,test/val_iceplains/3/000048.mp4,test/val_iceplains/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +180,180,398,test/val_plains/3/000010.mp4,test/val_plains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +181,181,435,test/val_plains/2/000027.mp4,test/val_plains/2/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +182,182,242,test/val_savanna/2/000049.mp4,test/val_savanna/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +183,183,361,test/val_iceplains/2/000045.mp4,test/val_iceplains/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +184,184,470,test/val_plains/2/000047.mp4,test/val_plains/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +185,185,356,test/val_iceplains/2/000042.mp4,test/val_iceplains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +186,186,468,test/val_plains/2/000046.mp4,test/val_plains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +187,187,113,test/val_desert/0/000043.mp4,test/val_desert/0/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +188,188,19,test/val_desert/1/000008.mp4,test/val_desert/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +189,189,199,test/val_savanna/0/000033.mp4,test/val_savanna/0/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +190,190,118,test/val_desert/2/000045.mp4,test/val_desert/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +191,191,433,test/val_plains/0/000027.mp4,test/val_plains/0/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +192,192,328,test/val_iceplains/0/000034.mp4,test/val_iceplains/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +193,193,213,test/val_savanna/0/000039.mp4,test/val_savanna/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +194,194,339,test/val_iceplains/0/000037.mp4,test/val_iceplains/0/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +195,195,466,test/val_plains/1/000046.mp4,test/val_plains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +196,196,355,test/val_iceplains/3/000042.mp4,test/val_iceplains/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +197,197,336,test/val_iceplains/3/000036.mp4,test/val_iceplains/3/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +198,198,420,test/val_plains/0/000020.mp4,test/val_plains/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +199,199,410,test/val_plains/1/000015.mp4,test/val_plains/1/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +200,200,27,test/val_desert/1/000011.mp4,test/val_desert/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +201,201,86,test/val_desert/1/000033.mp4,test/val_desert/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +202,202,210,test/val_savanna/3/000038.mp4,test/val_savanna/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +203,203,103,test/val_desert/3/000039.mp4,test/val_desert/3/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +204,204,245,test/val_iceplains/3/000000.mp4,test/val_iceplains/3/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +205,205,25,test/val_desert/2/000009.mp4,test/val_desert/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +206,206,340,test/val_iceplains/2/000037.mp4,test/val_iceplains/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +207,207,167,test/val_savanna/0/000016.mp4,test/val_savanna/0/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +208,208,424,test/val_plains/0/000022.mp4,test/val_plains/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +209,209,448,test/val_plains/2/000033.mp4,test/val_plains/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +210,210,386,test/val_plains/0/000006.mp4,test/val_plains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +211,211,392,test/val_plains/3/000008.mp4,test/val_plains/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +212,212,105,test/val_desert/1/000040.mp4,test/val_desert/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +213,213,267,test/val_iceplains/2/000008.mp4,test/val_iceplains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +214,214,90,test/val_desert/0/000034.mp4,test/val_desert/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +215,215,426,test/val_plains/0/000023.mp4,test/val_plains/0/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +216,216,168,test/val_savanna/3/000016.mp4,test/val_savanna/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +217,217,218,test/val_savanna/3/000040.mp4,test/val_savanna/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +218,218,445,test/val_plains/0/000032.mp4,test/val_plains/0/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +219,219,48,test/val_desert/3/000019.mp4,test/val_desert/3/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +220,220,104,test/val_desert/2/000039.mp4,test/val_desert/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +221,221,206,test/val_savanna/2/000035.mp4,test/val_savanna/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +222,222,228,test/val_savanna/1/000045.mp4,test/val_savanna/1/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +223,223,200,test/val_savanna/2/000033.mp4,test/val_savanna/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +224,224,62,test/val_desert/2/000024.mp4,test/val_desert/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +225,225,121,test/val_desert/1/000047.mp4,test/val_desert/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +226,226,147,test/val_savanna/0/000007.mp4,test/val_savanna/0/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +227,227,146,test/val_savanna/2/000006.mp4,test/val_savanna/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +228,228,225,test/val_savanna/1/000043.mp4,test/val_savanna/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +229,229,313,test/val_iceplains/3/000027.mp4,test/val_iceplains/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +230,230,354,test/val_iceplains/0/000042.mp4,test/val_iceplains/0/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +231,231,367,test/val_iceplains/1/000048.mp4,test/val_iceplains/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +232,232,297,test/val_iceplains/3/000020.mp4,test/val_iceplains/3/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +233,233,124,test/val_desert/1/000048.mp4,test/val_desert/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +234,234,171,test/val_savanna/2/000017.mp4,test/val_savanna/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +235,235,342,test/val_iceplains/0/000038.mp4,test/val_iceplains/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +236,236,60,test/val_desert/3/000023.mp4,test/val_desert/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +237,237,131,test/val_savanna/1/000001.mp4,test/val_savanna/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +238,238,365,test/val_iceplains/1/000047.mp4,test/val_iceplains/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +239,239,315,test/val_iceplains/2/000028.mp4,test/val_iceplains/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +240,240,220,test/val_savanna/1/000041.mp4,test/val_savanna/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +241,241,8,test/val_desert/1/000003.mp4,test/val_desert/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +242,242,261,test/val_iceplains/0/000006.mp4,test/val_iceplains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +243,243,17,test/val_desert/1/000007.mp4,test/val_desert/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +244,244,462,test/val_plains/1/000043.mp4,test/val_plains/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +245,245,454,test/val_plains/3/000038.mp4,test/val_plains/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +246,246,351,test/val_iceplains/1/000041.mp4,test/val_iceplains/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +247,247,383,test/val_plains/3/000005.mp4,test/val_plains/3/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +248,248,18,test/val_desert/2/000007.mp4,test/val_desert/2/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +249,249,457,test/val_plains/1/000040.mp4,test/val_plains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +250,250,326,test/val_iceplains/3/000033.mp4,test/val_iceplains/3/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +251,251,254,test/val_iceplains/1/000004.mp4,test/val_iceplains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +252,252,375,test/val_plains/1/000002.mp4,test/val_plains/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +253,253,141,test/val_savanna/0/000005.mp4,test/val_savanna/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +254,254,306,test/val_iceplains/2/000024.mp4,test/val_iceplains/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +255,255,10,test/val_desert/2/000003.mp4,test/val_desert/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +256,256,173,test/val_savanna/0/000020.mp4,test/val_savanna/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +257,257,320,test/val_iceplains/1/000032.mp4,test/val_iceplains/1/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +258,258,115,test/val_desert/1/000044.mp4,test/val_desert/1/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +259,259,341,test/val_iceplains/1/000038.mp4,test/val_iceplains/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +260,260,61,test/val_desert/3/000024.mp4,test/val_desert/3/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +261,261,231,test/val_savanna/1/000046.mp4,test/val_savanna/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +262,262,224,test/val_savanna/3/000042.mp4,test/val_savanna/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +263,263,292,test/val_iceplains/3/000018.mp4,test/val_iceplains/3/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +264,264,334,test/val_iceplains/1/000036.mp4,test/val_iceplains/1/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +265,265,159,test/val_savanna/0/000013.mp4,test/val_savanna/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +266,266,102,test/val_desert/0/000039.mp4,test/val_desert/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +267,267,362,test/val_iceplains/1/000046.mp4,test/val_iceplains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +268,268,322,test/val_iceplains/3/000032.mp4,test/val_iceplains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +269,269,381,test/val_plains/1/000004.mp4,test/val_plains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +270,270,84,test/val_desert/3/000032.mp4,test/val_desert/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +271,271,243,test/val_iceplains/1/000000.mp4,test/val_iceplains/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +272,272,65,test/val_desert/3/000025.mp4,test/val_desert/3/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +273,273,64,test/val_desert/0/000025.mp4,test/val_desert/0/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +274,274,272,test/val_iceplains/2/000010.mp4,test/val_iceplains/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +275,275,399,test/val_plains/0/000011.mp4,test/val_plains/0/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +276,276,248,test/val_iceplains/3/000001.mp4,test/val_iceplains/3/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +277,277,43,test/val_desert/3/000017.mp4,test/val_desert/3/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +278,278,393,test/val_plains/2/000008.mp4,test/val_plains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +279,279,123,test/val_desert/2/000047.mp4,test/val_desert/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +280,280,29,test/val_desert/2/000011.mp4,test/val_desert/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +281,281,122,test/val_desert/0/000047.mp4,test/val_desert/0/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +282,282,157,test/val_savanna/3/000011.mp4,test/val_savanna/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +283,283,333,test/val_iceplains/2/000035.mp4,test/val_iceplains/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +284,284,307,test/val_iceplains/1/000025.mp4,test/val_iceplains/1/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +285,285,9,test/val_desert/0/000003.mp4,test/val_desert/0/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +286,286,250,test/val_iceplains/2/000002.mp4,test/val_iceplains/2/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +287,287,165,test/val_savanna/3/000015.mp4,test/val_savanna/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +288,288,0,test/val_desert/1/000000.mp4,test/val_desert/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +289,289,271,test/val_iceplains/3/000010.mp4,test/val_iceplains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +290,290,276,test/val_iceplains/1/000013.mp4,test/val_iceplains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +291,291,82,test/val_desert/1/000031.mp4,test/val_desert/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +292,292,191,test/val_savanna/2/000029.mp4,test/val_savanna/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +293,293,374,test/val_plains/1/000001.mp4,test/val_plains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +294,294,446,test/val_plains/3/000032.mp4,test/val_plains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +295,295,378,test/val_plains/1/000003.mp4,test/val_plains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +296,296,152,test/val_savanna/0/000009.mp4,test/val_savanna/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +297,297,54,test/val_desert/0/000021.mp4,test/val_desert/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +298,298,259,test/val_iceplains/2/000005.mp4,test/val_iceplains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +299,299,91,test/val_desert/2/000034.mp4,test/val_desert/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.sha256 b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..065350069e5b0f46e33cd51ad63f60aa956822fd --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.sha256 @@ -0,0 +1 @@ +ef4e4de9d3975d8a7e724bb609e57f20186cd1f7285457c00fb85c1aca973b99 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_0.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_0.log new file mode 100644 index 0000000000000000000000000000000000000000..eeba67515f2330016b71fc909c4cf895aec978ed --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_0.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:32,957][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:41,370][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:41,520][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:41,520][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:41,520][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:41,520][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,323][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=0', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=0', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_1.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_1.log new file mode 100644 index 0000000000000000000000000000000000000000..8ca9d635ed14164fe4858c6f994bb8a59b6b22ae --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_1.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:20,583][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:30,677][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:30,833][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:30,833][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:30,834][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:30,834][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:47,563][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=10', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=10', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_10.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_10.log new file mode 100644 index 0000000000000000000000000000000000000000..8a9ed3df3462c4ddb3319b7a6a3f2f164755e2f1 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_10.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:25,008][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:33,433][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:33,583][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:33,584][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:33,584][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:33,584][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:48,171][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=100', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=100', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_11.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_11.log new file mode 100644 index 0000000000000000000000000000000000000000..bbc9c922bdf35d3d0f6ed4310b688fb65a1cd245 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_11.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:20,456][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:32,148][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,308][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,308][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,308][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,308][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:48,186][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=110', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=110', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_12.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_12.log new file mode 100644 index 0000000000000000000000000000000000000000..124e7445fb89be32892770fde8bf1215b4f527ae --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_12.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:20,512][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:31,454][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:31,759][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:31,760][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:31,760][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:31,760][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:48,650][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=120', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=120', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_13.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_13.log new file mode 100644 index 0000000000000000000000000000000000000000..53c1b45d8150e097de62f1574cf6436f7ebd5ed8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_13.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:20,841][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:32,162][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,319][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,319][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,319][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,319][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:48,669][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=129', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=129', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_14.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_14.log new file mode 100644 index 0000000000000000000000000000000000000000..f3e842739dc14aa6ce37e0de1aa85215fa813128 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_14.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:26,033][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:34,392][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:34,550][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:34,550][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:34,550][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:34,550][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,207][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=138', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=138', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_15.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_15.log new file mode 100644 index 0000000000000000000000000000000000000000..9214c987b3879903fceab15c8a6f623afb6f7354 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_15.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:32,391][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:40,748][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:40,889][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:40,889][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:40,889][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:40,889][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,314][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=147', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=147', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_16.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_16.log new file mode 100644 index 0000000000000000000000000000000000000000..5fea4bab691b79c8e64610ee1a9592e2f7e872a1 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_16.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:25,647][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:34,095][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:34,240][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:34,240][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:34,240][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:34,240][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,183][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=156', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=156', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_17.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_17.log new file mode 100644 index 0000000000000000000000000000000000000000..723d2e6a82a0ab6efe1e4f7c8f493369686c4a0a --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_17.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:23,784][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:32,242][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,398][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,398][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,399][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,399][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:45,598][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=165', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=165', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_18.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_18.log new file mode 100644 index 0000000000000000000000000000000000000000..bf671cd7522d46608985098130840d12dfd888e8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_18.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:25,718][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:34,173][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:34,318][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:34,318][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:34,318][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:34,318][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,160][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=174', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=174', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_19.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_19.log new file mode 100644 index 0000000000000000000000000000000000000000..1dfbd773bbe9a6fc12a494dd1b4fa1727a30b3ca --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_19.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:20,460][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:30,677][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:30,833][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:30,833][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:30,833][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:30,833][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:46,451][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=183', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=183', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_2.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_2.log new file mode 100644 index 0000000000000000000000000000000000000000..8fb3f2b505620f7eca665bbd7a724ffcfbba892d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_2.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:22,146][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:31,445][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:31,881][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:31,882][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:31,882][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:31,882][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:48,661][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=20', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=20', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_20.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_20.log new file mode 100644 index 0000000000000000000000000000000000000000..a236f8b265380e5ee77982a52d1e5dd315760602 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_20.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:24,680][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:33,059][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:33,208][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:33,208][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:33,208][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:33,208][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:44,659][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=192', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=192', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_21.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_21.log new file mode 100644 index 0000000000000000000000000000000000000000..b557347029b490501931c2622832762c01088678 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_21.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:20,151][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:30,679][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:31,512][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:31,512][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:31,512][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:31,512][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:47,578][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=201', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=201', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_22.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_22.log new file mode 100644 index 0000000000000000000000000000000000000000..bec3b679a42be171c545b49695e936df09bf71f9 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_22.log @@ -0,0 +1,66 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:20,626][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:30,678][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,087][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,087][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,087][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,087][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +WARNING: Missing logger folder: /proj/cvl/users/x_fahkh2/DeMemWM/lightning_logs +[2026-07-22 14:02:43,124][lightning.fabric.loggers.csv_logs][WARNING] - Missing logger folder: /proj/cvl/users/x_fahkh2/DeMemWM/lightning_logs +[2026-07-22 14:02:43,129][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=210', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=210', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_23.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_23.log new file mode 100644 index 0000000000000000000000000000000000000000..f46ecc95f15d4665f5d74c4df3e275a3485e03fb --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_23.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:26,691][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:35,151][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:35,299][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:35,299][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:35,299][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:35,299][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,241][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=219', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=219', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_24.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_24.log new file mode 100644 index 0000000000000000000000000000000000000000..b99dc03f8de07e4dc9150393232e67f3f6e07d87 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_24.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:20,429][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:32,413][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,561][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,561][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,561][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,561][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,138][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=228', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=228', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_25.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_25.log new file mode 100644 index 0000000000000000000000000000000000000000..a4f12b752069ec37078bb03fdbb68360ee3806aa --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_25.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:28,356][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:36,775][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:36,927][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:36,927][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:36,927][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:36,927][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,301][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=237', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=237', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_26.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_26.log new file mode 100644 index 0000000000000000000000000000000000000000..2dd36789959beadf18a810ab0e2064c305d5fa44 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_26.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:25,902][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:34,305][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:34,448][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:34,448][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:34,448][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:34,448][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,194][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=246', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=246', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_27.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_27.log new file mode 100644 index 0000000000000000000000000000000000000000..b903aac943b89922729e6127df6ccd41932152c3 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_27.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:26,683][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:35,525][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:35,685][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:35,686][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:35,686][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:35,686][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,263][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=255', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=255', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_28.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_28.log new file mode 100644 index 0000000000000000000000000000000000000000..0bbf62cb3c55dcd4763e072ffd8d729f0859e13d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_28.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:27,342][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:35,793][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:35,936][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:35,937][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:35,937][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:35,937][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,277][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=264', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=264', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_29.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_29.log new file mode 100644 index 0000000000000000000000000000000000000000..37c61c39473418a1d77ae3472bdf8d12454916e2 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_29.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:20,039][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:32,419][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:33,334][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:33,334][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:33,334][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:33,334][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:46,391][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=273', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=273', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_3.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_3.log new file mode 100644 index 0000000000000000000000000000000000000000..a77fe707dfbbd3f2305ac012b871d04474154bee --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_3.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:26,089][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:34,528][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:34,673][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:34,673][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:34,673][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:34,673][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,170][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=30', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=30', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_30.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_30.log new file mode 100644 index 0000000000000000000000000000000000000000..d82c7f438c901601a72a50e94efbad1f4c74d044 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_30.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:26,495][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:34,892][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:35,050][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:35,050][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:35,050][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:35,050][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,253][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=282', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=282', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_31.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_31.log new file mode 100644 index 0000000000000000000000000000000000000000..a56722a843ce6309fca9714287555fd4ac2738f5 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_31.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:24,411][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:32,817][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,963][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,963][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,963][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,963][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:47,656][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=291', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=291', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_4.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_4.log new file mode 100644 index 0000000000000000000000000000000000000000..803c79f06718a853417f1aa3f6af1dddc0fa45cd --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_4.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:20,130][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:32,454][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,603][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,603][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,603][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,603][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:46,709][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=40', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=40', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_5.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_5.log new file mode 100644 index 0000000000000000000000000000000000000000..e4b384d176040e28fcb36e401b2ddefeff294d95 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_5.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:25,839][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:34,353][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:34,499][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:34,499][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:34,499][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:34,499][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,230][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=50', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=50', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_6.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_6.log new file mode 100644 index 0000000000000000000000000000000000000000..bf242d64466d71392d1b596b36627fbc0f4f6678 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_6.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:29,096][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:37,514][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:37,671][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:37,671][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:37,671][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:37,671][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,288][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=60', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=60', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_7.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_7.log new file mode 100644 index 0000000000000000000000000000000000000000..5b9a9a1a9475266b780dfee2ae4f312c31a2f677 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_7.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:20,542][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:32,174][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,336][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,337][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,337][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,337][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:44,298][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=70', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=70', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_8.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_8.log new file mode 100644 index 0000000000000000000000000000000000000000..f4305575d33bffbe7317d4d5c063b846341e8411 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_8.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:25,413][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:34,012][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:34,158][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:34,158][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:34,158][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:34,159][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,217][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=80', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=80', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_9.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_9.log new file mode 100644 index 0000000000000000000000000000000000000000..ee444773f797dd8753dd9ba291b5dd82bf966617 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/worker_9.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 42 +[2026-07-22 14:02:24,219][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:02:32,683][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,825][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,825][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,825][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,825][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:48,679][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=42', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=90', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=42', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=90', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..31f11dc225a3082fbbd8f02f36e10ce9006ee12c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 0 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 0 + eval_manifest_count: 10 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..1b43fb777a7fe9cd4ddf4342037cf9f63bfe7ffe --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=0 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=0 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=0,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=0,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b41377f516f4042c876fbf3edd83afdfca8f7a23 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=0 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=0 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..a4f12f26a6f49b525388768aef054b0ba49a3c9e --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_00/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:32,957][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:41,370][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:41,520][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:41,520][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:41,520][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:41,520][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,323][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cbbad7d136841d2c873482c46f716fd3746af528 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 10 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 10 + eval_manifest_count: 10 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cf9708ca17df6c769d933226a2574ad37415a42d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=10 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=10 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=10,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=10,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3c08078730b50271763448dfdcc8932def2b34f6 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=10 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=10 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..75716ddacca4fca5faee2257e6310c93536855d6 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_01/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:20,583][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:30,677][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:30,833][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:30,833][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:30,834][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:30,834][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:47,563][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..87e44f5c1ab437b588d517bcfe19fe2c4307c755 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 20 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 20 + eval_manifest_count: 10 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f4c9e3d1722ff9e65ae54215cbecb78452b6791d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=20 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=20 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=20,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=20,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..2d0381ef32dcc799409ff1d4168d0a528ea2cfcd --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=20 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=20 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..7b072ac762ae6fc2580c845c0bdd023439ef1140 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_02/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:22,146][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:31,445][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:31,881][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:31,882][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:31,882][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:31,882][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:48,661][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d4904e1bf47021c7ce6730ecbfc7af452ca49555 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 30 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 30 + eval_manifest_count: 10 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d2f3f902e388624b15cf995b8b6221eece65ba53 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=30 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=30 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=30,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=30,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ced10f1266265d96196a5a1e6254013a0a458ef8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=30 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=30 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..96ed3dfa47457ba1ba2e958ece79e4bcc96e8faf --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_03/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:26,089][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:34,528][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:34,673][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:34,673][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:34,673][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:34,673][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,170][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75ef28e693ceea98178d42166a1a4efeb7935278 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 40 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 40 + eval_manifest_count: 10 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4d11ab79d04463a710cb22f33d400e9df3d1dc6a --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=40 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=40 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=40,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=40,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..77f1428cf4fcfa320ee47e4d7d15aaa60922c104 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=40 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=40 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..8cb9512d55a2de61a23b886d10c6d7b443e12ca8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_04/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:20,130][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:32,454][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,603][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,603][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,603][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,603][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:46,709][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..aeb3393ba9d832bd6cf569424c233d957f325331 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 50 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 50 + eval_manifest_count: 10 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b9a61c5ca8a358026ba12cb5166aae5bd43d6ace --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=50 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=50 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=50,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=50,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3a641c1724a386ae300930757df72465547d3f62 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=50 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=50 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..9a877c9083cf1ddc7f629f469e17f418ce71a2de --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_05/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:25,839][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:34,353][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:34,499][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:34,499][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:34,499][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:34,499][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,230][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3819aac832e12410de6ee95c75171e7f5c8c5503 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 60 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 60 + eval_manifest_count: 10 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ced7af88a24ddf3f48e927aa6c66f749d8c3b613 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=60 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=60 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=60,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=60,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a5260c2979288d801817975408bfc3f545ea9773 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=60 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=60 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..ae2b31a2d2f4b3ae55741e53e8f4adeea780281e --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_06/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:29,096][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:37,514][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:37,671][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:37,671][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:37,671][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:37,671][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,288][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..90e7520f6415fd473fe853dc52377e05e46062cc --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 70 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 70 + eval_manifest_count: 10 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..611ba2917dc4d008b1decdb1196c7826582bb009 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=70 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=70 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=70,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=70,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..00b065acf22c25ca8f05f7a9118af0ec9e3860e8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=70 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=70 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..84161f2494958e0f1bfaeed68e64dc75dee8d767 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_07/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:20,542][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:32,174][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,336][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,337][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,337][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,337][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:44,298][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..751eb447c9158f7b49c0723b06abc58497c01d62 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 80 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 80 + eval_manifest_count: 10 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..74a98481fe2198fc926551e2ba18a4ce7fe801da --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=80 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=80 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=80,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=80,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..dfde96ad782733f2207bd15fcd0d42a3fe5d3215 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=80 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=80 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..9252e3518b75ad73204b830bf04e81c1c60f1a45 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_08/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:25,413][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:34,012][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:34,158][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:34,158][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:34,158][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:34,159][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,217][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9eadaf7c8a596b9dcd9e8acb7020c562097641b7 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 90 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 90 + eval_manifest_count: 10 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..300eae9c5d977bce805b4f55fcb08c23345d2000 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=90 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=90 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=90,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=90,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..81ae5c1697867a9952b0bdf0a0f501b13ea2959c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=90 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=90 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e904e9245d05a0046d1cfaaa1fb25f13d4ef253d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_09/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:24,219][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:32,683][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,825][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,825][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,825][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,825][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:48,679][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4f4e6050ad877190d26c5729ee787375c80a453c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 100 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 100 + eval_manifest_count: 10 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..372cf548a2b02c94f7728925d3f2c54ac6419482 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=100 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=100 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=100,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=100,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b1702d6a2a34cf00b27630b3e4f57daf74147d2d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=100 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=100 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..d34f9ab34cd398a7b3001410bfee53965abeb816 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_10/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:25,008][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:33,433][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:33,583][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:33,584][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:33,584][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:33,584][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:48,171][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..171842d3fa2b5a6c4c645c13a2e95b1a917a7818 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 110 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 110 + eval_manifest_count: 10 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..22a5d4e675aff0adfb36786ea33e40a13ef5e284 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=110 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=110 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=110,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=110,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..229cdda928ff050cbea44c6eeb434886ae931129 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=110 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=110 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..d9d508b85e285b4d6f3efbe902be873d1862ca8d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_11/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:20,456][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:32,148][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,308][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,308][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,308][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,308][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:48,186][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..228b023e6fef28e8e634cfca6d942deffb689447 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 120 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 120 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f95e4f4c8ccfed2e04b18e51bf27a4de30d194c9 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=120 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=120 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=120,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=120,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3e2b988d22dbf1c1187c185fba6905d61d0faccc --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=120 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=120 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..1d71715d436da08ab98631091746c54eb7d111d5 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_12/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:20,512][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:31,454][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:31,759][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:31,760][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:31,760][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:31,760][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:48,650][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..dd6a92908215135d3a9f11ffbeaffba211d828ce --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 129 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 129 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8f8488987e90d4f4f43c6881013da6dec564cec9 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=129 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=129 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=129,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=129,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..67d7b3645a55fb7fecbbfe3143fe57d311161ddc --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=129 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=129 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..63a977fb78e3a63a7c4b537a2e31954ca843d200 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_13/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:20,841][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:32,162][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,319][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,319][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,319][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,319][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:48,669][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4ce09cb8ea628b08b0615a7e4f67588a6468c789 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 138 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 138 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..72ddc8ef9231354263b21d6e02090fa070b4470a --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=138 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=138 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=138,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=138,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..1ec402072cb7ad8986906206a0244429b7318ffb --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=138 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=138 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..7af0efce70d402f3e866f2cc4b904fda6f061c1b --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_14/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:26,033][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:34,392][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:34,550][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:34,550][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:34,550][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:34,550][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,207][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9a944f306af697465e5d09f56f405743f1216ab5 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 147 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 147 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..476e6141c096e64078d1babb0be92897432e154d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=147 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=147 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=147,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=147,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..1f02cdeaf791f151bb0712fd7068595e5318934a --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=147 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=147 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..42bd50e0b6efca2d9fea90911665f5ef1b8427b7 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_15/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:32,391][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:40,748][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:40,889][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:40,889][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:40,889][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:40,889][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,314][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b230b3176a6d637561c440f4c400b051bdef5ebb --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 156 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 156 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..43cb63c8235c2f13902aac0fa5870fb7adbb61fc --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=156 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=156 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=156,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=156,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d87687094f75c17d3a7c2f4955ff869cdefaa7c9 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=156 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=156 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..ccb281d05206932acb33b331910cea9f1f3ea90c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_16/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:25,647][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:34,095][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:34,240][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:34,240][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:34,240][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:34,240][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,183][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c69dcaf6cc82592198a1b43deb4e266c94cafb24 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 165 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 165 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ddbd2ab589885d01245f803bc82f697b63b931c5 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=165 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=165 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=165,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=165,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a659f33e200a2b2fc5f8196a8a8a7a7788912a1d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=165 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=165 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..40271d4da10f7b870e80fd5719e5537e841ae6b5 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_17/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:23,784][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:32,242][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,398][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,398][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,399][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,399][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:45,598][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..110595fb347f1d723aca307b7a187329a9122187 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 174 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 174 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3fe3dbb262c1f1457cc2164a859d2d3e7307ee5b --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=174 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=174 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=174,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=174,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c0adfd7b8a8099c9841b7b170feedb6e563b3e49 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=174 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=174 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..4c61355b986601f996b3030b5d4a395b780af3bf --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_18/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:25,718][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:34,173][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:34,318][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:34,318][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:34,318][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:34,318][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,160][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..2de09dd9a5a702db765c93e3dbb1a894fc31e9e1 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 183 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 183 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a6b4eb3843ac126426330de3959bc2f129071b63 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=183 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=183 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=183,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=183,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..867238aac226449f00fc011e3fe16786d481956e --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=183 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=183 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..bcb3c25f6cdb0a02f5f269e96edd0fd66dd0648a --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_19/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:20,460][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:30,677][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:30,833][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:30,833][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:30,833][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:30,833][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:46,451][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b8f8eca7a8288172a7b820a35b7c37f3be0ea76d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 192 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 192 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..20bdef3746e378a742cf80345887ea1735fb6376 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=192 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=192 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=192,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=192,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..57a065f09b62a7f0928e48d1437165d1105d0457 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=192 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=192 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..4a7b6b3d45b8f2cd97308595e6436406ef309cde --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_20/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:24,680][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:33,059][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:33,208][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:33,208][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:33,208][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:33,208][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:44,659][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..42789f6eeef97028a143ee1c34b3ccaec514b96e --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 201 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 201 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..55518ef5198914a6e729a8aabc324b7552a59e5c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=201 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=201 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=201,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=201,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..af4f3ae0a9191ac956ca874bd26b72257e1cd797 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=201 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=201 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..cc5feed244570af82c01a2eddd30ccce0a7883a5 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_21/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:20,151][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:30,679][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:31,512][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:31,512][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:31,512][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:31,512][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:47,578][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..dcb1aa7f144702e05d16bd25333347c3df09fd87 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 210 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 210 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..95edfc432c3f06f8528653051d93185a9f56ee9e --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=210 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=210 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=210,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=210,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0174a39e92d653c708e3661b1bc3c57cb3007188 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=210 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=210 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..77fe3d9989f9585e95227c5b3b2e0df50d54c6e2 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_22/hydra/run/main.log @@ -0,0 +1,8 @@ +[2026-07-22 14:02:20,626][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:30,678][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,087][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,087][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,087][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,087][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:43,124][lightning.fabric.loggers.csv_logs][WARNING] - Missing logger folder: /proj/cvl/users/x_fahkh2/DeMemWM/lightning_logs +[2026-07-22 14:02:43,129][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c99bb3f34c384c277f69c4f1e81376e978181e25 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 219 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 219 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..35f75ba01df6384d00e8ffa7674e46fd5ff2b76f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=219 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=219 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=219,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=219,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ca3f6bd2ff8b5fb5cac36b89c0648992df2c208d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=219 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=219 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..517532a9227b898e52467aa550f73b83f8af6d1d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_23/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:26,691][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:35,151][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:35,299][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:35,299][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:35,299][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:35,299][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,241][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e7537e9ae17ebe8661665e0a0e3468d4d46c2e50 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 228 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 228 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c6c80d41d9c438050c6132368b53048158a9f71d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=228 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=228 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=228,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=228,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..95a8240bb20b79125b9a40ec4b7a7f04f3756349 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=228 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=228 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..eb1ae17901ea2b2d1ce78d1bec48e72ddcb05489 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_24/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:20,429][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:32,413][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,561][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,561][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,561][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,561][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,138][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fe4a8e7f5ed81f8c2120a28848e6661c0b02c06e --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 237 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 237 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f691531461918bfe4346eb08dca41853956cb075 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=237 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=237 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=237,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=237,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7c169c38376b401485f964a887f7bf4510168f0b --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=237 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=237 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..6e503ab86c055746c6947712b3e2ebeeee946857 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_25/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:28,356][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:36,775][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:36,927][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:36,927][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:36,927][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:36,927][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,301][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c0957641b1c9e3b4facde88f29061962b1cea17a --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 246 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 246 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9e7a74213abf02c8687b5900d0f912378dea51ae --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=246 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=246 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=246,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=246,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9f2dc229e0563fe14fa4f66c0ac821ea8a842ce1 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=246 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=246 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..62056558beddea9583bcee4b40bb9051cd153831 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_26/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:25,902][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:34,305][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:34,448][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:34,448][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:34,448][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:34,448][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,194][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d49c385a1e0fecea8b25a92e30b67071fb642788 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 255 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 255 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..39573f79955f31b792f84185773384b4aa6f46d0 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=255 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=255 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=255,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=255,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4b60aabdc151120d9e3d2a41e269843742929b0c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=255 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=255 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..c49f5798f5571ffb435225ea165f67e88f5ec42a --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_27/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:26,683][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:35,525][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:35,685][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:35,686][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:35,686][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:35,686][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,263][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..00cd203027fd2000ac674543244616ebfdf94fcf --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 264 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 264 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cfc64a76f865e937c9af6214f00ca5eb13ba6849 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=264 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=264 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=264,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=264,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..208a2559233e373f97099efafc9de194395584bb --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=264 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=264 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..9b254d4b3ffe6561b3faf0e6b517907e41a23487 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_28/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:27,342][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:35,793][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:35,936][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:35,937][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:35,937][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:35,937][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,277][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b00c82f8203e58850a267910e28d72b4cc503ca8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 273 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 273 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9b18c77c0744a48cf01b5f579e41ff4eb2ffcefe --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=273 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=273 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=273,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=273,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a6fa86b13f907ace5257d85c44bf5644c98c9a23 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=273 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=273 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..0fc1df71c1e9ee47a8e08483c8b7d2235766c26b --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_29/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:20,039][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:32,419][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:33,334][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:33,334][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:33,334][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:33,334][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:46,391][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8b3962c79cec53ad667f0919495639cc98ecfee7 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 282 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 282 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8e57e8518901b1f0a01753d735f8a637870c1836 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=282 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=282 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=282,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=282,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b51b5a286b8fc8af9b6c6c0b2cf4275a111261ea --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=282 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=282 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..1f9c3a8a5299377715ff75e0ef7889a67cf06ce3 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_30/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:26,495][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:34,892][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:35,050][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:35,050][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:35,050][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:35,050][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:49,253][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..52b8ab54097897cb6f956e8754a08af402d28405 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 291 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 291 + eval_manifest_count: 9 + eval_master_seed: 42 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +customized_load: false +seed: 42 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f57a9e5e5be97f4f0ce1cdd73285ea8d111b422d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=42 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=291 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=42 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=291 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=291,+algorithm.eval_master_seed=42,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=291,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42,+seed=42,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ebf3a7277c1bd5e42a862781621ec247813bbcd4 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=42 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=291 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=42 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=291 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..10b9144d5b450b0f4ddb5e3744042dfb83a1cffd --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/workers/worker_31/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:02:24,411][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 42 +[2026-07-22 14:02:32,817][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:02:32,963][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:02:32,963][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:02:32,963][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:02:32,963][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:02:47,656][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/checkpoint_sha256.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/checkpoint_sha256.csv new file mode 100644 index 0000000000000000000000000000000000000000..0cbabb25f31be55688d79358c5cfed3fa5754432 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/checkpoint_sha256.csv @@ -0,0 +1,2 @@ +path,size_bytes,sha256 +/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,11853520525,c33c96daf31163d9efcdcf800479ad6b599353fdd81929b1ebd10af9044e75ec diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/evaluation.patch b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/evaluation.patch new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..3cfd08bc49f4159d2bd5cace6409ac92956d78e1 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.csv @@ -0,0 +1,51 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.sha256 b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..67b0dc2042229c0768d324282215bdf64bebe28f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.sha256 @@ -0,0 +1 @@ +f9536911e434610cce23306ff99575db53cd7fc29e3df841da4b6631bf418634 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/git_provenance.json b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/git_provenance.json new file mode 100644 index 0000000000000000000000000000000000000000..fabe691fc08806f957f2445d4a52b2f38d7f07d8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/git_provenance.json @@ -0,0 +1,5 @@ +{ + "branch": "dememwm-unified-sourcefree-age-balanced", + "head": "732555281e86c4fdbb71c98a000d648c3edc3109", + "status": "?? results/\n?? train_dememwm_full_berzelius_gate_off_h200.sh\n?? train_dememwm_separate_mem_gatefix_4layers512_valid500_berzelius_h200.sh" +} diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/launcher.sh b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/launcher.sh new file mode 100644 index 0000000000000000000000000000000000000000..db637f1236690823d28fd0f38aa79441ca248693 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/launcher.sh @@ -0,0 +1,182 @@ +#!/usr/bin/env bash +#SBATCH --job-name=demem-sf-countoff-lh300 +#SBATCH --nodes=1 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=14 +#SBATCH --gres=gpu:8 +#SBATCH --time=48:00:00 +#SBATCH --account=berzelius-2026-196 +#SBATCH --array=42-43%1 +#SBATCH --chdir=/proj/cvl/users/x_fahkh2/DeMemWM +#SBATCH --output=/proj/cvl/users/x_fahkh2/caches/slurm-%x-%A_%a.out +#SBATCH --error=/proj/cvl/users/x_fahkh2/caches/slurm-%x-%A_%a.out + +set -eo pipefail + +module load buildenv-gcccuda/12.9.1-gcc11 +source $(conda info --base)/etc/profile.d/conda.sh + +REPO=/proj/cvl/users/x_fahkh2/DeMemWM +DATA_DIR=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +FEATURE_DIR=$DATA_DIR/vae_features +CACHE_DIR=/proj/cvl/users/x_fahkh2/caches +TRAIN_RUN=train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k +MODEL=dememwm_sourcefree_countcorr_off_step200k +CHECKPOINT=$REPO/results/$TRAIN_RUN/checkpoints/epoch3_step200000.ckpt +OUTPUT_ROOT=$REPO/outputs/dememwm_sourcefree_long_horizon +MANIFEST=$REPO/evaluation/manifests/minecraft_test_300.csv +FID_MANIFEST=$REPO/evaluation/manifests/minecraft_test_fid50.csv +FID_BATCH_SIZE=32 + +export PYTHONPATH="$REPO:${PYTHONPATH:-}" +export HF_HOME=$CACHE_DIR +export TORCH_HOME=$CACHE_DIR +export PIP_CACHE_DIR=$CACHE_DIR +export TMPDIR=/scratch/local/$USER/dememwm_eval_$SLURM_JOB_ID +export TRITON_CACHE_DIR=$TMPDIR +export CUDA_HOME=$CUDA_ROOT +export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +export WANDB_DISABLED=true +export HYDRA_FULL_ERROR=1 +export PYTHONWARNINGS=ignore +export PYTHONUNBUFFERED=1 +export OMP_NUM_THREADS=1 +export MKL_NUM_THREADS=1 + +cd "$REPO" +mkdir -p "$TMPDIR" "$OUTPUT_ROOT" + +MASTER_SEED=$SLURM_ARRAY_TASK_ID +RUN_NAME=${MODEL}_minecraft_test300_ctx600_roll500_bs1_seed${MASTER_SEED} +OUTPUT_DIR=$OUTPUT_ROOT/$RUN_NAME +WORKERS_DIR=$OUTPUT_DIR/workers +RANGES=( + 0:10 10:20 20:30 30:40 + 40:50 50:60 60:70 70:80 + 80:90 90:100 100:110 110:120 + 120:129 129:138 138:147 147:156 + 156:165 165:174 174:183 183:192 + 192:201 201:210 210:219 219:228 + 228:237 237:246 246:255 255:264 + 264:273 273:282 282:291 291:300 +) + +python long_horizon_eval.py verify-manifest --manifest "$MANIFEST" --clips 300 +python long_horizon_eval.py verify-manifest --manifest "$FID_MANIFEST" --clips 50 +python long_horizon_eval.py init-run \ + --output "$OUTPUT_DIR" \ + --repo "$REPO" \ + --manifest "$MANIFEST" \ + --fid-manifest "$FID_MANIFEST" \ + --model "$MODEL" \ + --master-seed "$MASTER_SEED" \ + --batch-size 1 \ + --checkpoint "$CHECKPOINT" \ + "${RANGES[@]/#/--worker-range=}" + +mkdir -p "$WORKERS_DIR" +cp evaluate_dememwm_sourcefree_countcorr_off_long_horizon_8gpu.sh "$OUTPUT_DIR/launcher.sh" + +PIDS=() +for WORKER_ID in "${!RANGES[@]}"; do + GPU_ID=$((WORKER_ID / 4)) + START=${RANGES[$WORKER_ID]%:*} + END=${RANGES[$WORKER_ID]#*:} + COUNT=$((END - START)) + WORKER_DIR=$(printf '%s/worker_%02d' "$WORKERS_DIR" "$WORKER_ID") + mkdir -p "$WORKER_DIR" + + CUDA_VISIBLE_DEVICES=$GPU_ID python -m main \ + +name="$RUN_NAME" \ + hydra.run.dir="$WORKER_DIR/hydra/run" \ + load="$CHECKPOINT" \ + auto_resume=false \ + +customized_load=false \ + wandb.mode=disabled \ + "experiment.tasks=[test]" \ + experiment.num_nodes=1 \ + +seed="$MASTER_SEED" \ + dataset=video_minecraft_dememwm_latent \ + dataset.save_dir="$DATA_DIR" \ + dataset.precomputed_feature_dir="$FEATURE_DIR" \ + dataset.n_frames=8 \ + dataset.n_frames_valid=1100 \ + dataset.context_length=100 \ + dataset.single_eval_clip=true \ + dataset.memory_selection.enabled=true \ + dataset.memory_selection.causal=true \ + dataset.memory_selection.max_anchor_frames=2 \ + dataset.memory_selection.max_dynamic_frames=4 \ + dataset.memory_selection.max_revisit_frames=2 \ + dataset.memory_selection.training_dropout=0.1 \ + dataset.memory_selection.dynamic.selection_policy=multiview \ + dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse \ + +dataset.seed=42 \ + +dataset.wo_updown=false \ + algorithm=dememwm_base \ + algorithm.causal=true \ + algorithm.n_tokens=8 \ + algorithm.context_frames=600 \ + algorithm.metric_report_segment=100 \ + algorithm.use_memory_attention=true \ + algorithm.anchor_zero_geometry=false \ + algorithm.memory_role_embedding=true \ + algorithm.memory_source_embedding=false \ + algorithm.memory_attention_key_only_geometry=true \ + algorithm.memory_age.enabled=true \ + algorithm.memory_age.num_buckets=32 \ + algorithm.memory_age.num_exact_buckets=17 \ + algorithm.memory_age.overflow_start=100 \ + algorithm.memory_stream_count_correction=false \ + algorithm.noise_route.anchor=high \ + algorithm.noise_route.dynamic=low \ + algorithm.noise_route.revisit=all \ + algorithm.memory_source_simulation.training_generated_suffix_max_frames=null \ + algorithm.memory_noise.enabled=true \ + algorithm.memory_noise.mode=source_free_logsnr \ + algorithm.memory_noise.stable_probability=0.25 \ + algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 \ + algorithm.memory_noise.validation_noisy_memory=false \ + algorithm.chunk_size=1 \ + algorithm.diffusion.sampling_timesteps=20 \ + algorithm.diffusion.ddim_sampling_eta=0.0 \ + algorithm.log_video=false \ + algorithm.save_local=true \ + +algorithm.local_save_dir="$WORKER_DIR" \ + +algorithm.lpips_batch_size=16 \ + +algorithm.eval_manifest_path="$MANIFEST" \ + +algorithm.eval_manifest_start="$START" \ + +algorithm.eval_manifest_count="$COUNT" \ + +algorithm.eval_master_seed="$MASTER_SEED" \ + +algorithm.eval_batch_size=1 \ + +experiment.test.manifest_path="$MANIFEST" \ + +experiment.test.manifest_start="$START" \ + +experiment.test.manifest_count="$COUNT" \ + experiment.test.batch_size=1 \ + experiment.test.limit_batch="$COUNT" \ + experiment.test.data.num_workers=2 \ + experiment.test.inference_mode=false \ + > "$OUTPUT_DIR/worker_${WORKER_ID}.log" 2>&1 & + PIDS+=("$!") +done + +FAILED=0 +for PID in "${PIDS[@]}"; do + wait "$PID" || FAILED=1 +done +test "$FAILED" -eq 0 + +python -c 'from omegaconf import OmegaConf; import sys; cfg=OmegaConf.load(sys.argv[1]); OmegaConf.save(cfg, sys.argv[2], resolve=True)' \ + "$WORKERS_DIR/worker_00/hydra/run/.hydra/config.yaml" "$OUTPUT_DIR/resolved_config.yaml" +python long_horizon_eval.py aggregate \ + --manifest "$MANIFEST" \ + --workers-root "$WORKERS_DIR" \ + --output "$OUTPUT_DIR" \ + --master-seed "$MASTER_SEED" \ + --workers 32 +srun --exclusive --exact --nodes=1 --ntasks=1 --gpus-per-task=1 --cpus-per-task=8 \ + python calculate_fid.py \ + --run-dir "$OUTPUT_DIR" \ + --cache-dir "$CACHE_DIR/fid_real_stats" \ + --batch-size "$FID_BATCH_SIZE" \ + --device cuda diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/protocol.json b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/protocol.json new file mode 100644 index 0000000000000000000000000000000000000000..c70560f87ff7c86553c451b824c2ae3a15b0f37a --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/protocol.json @@ -0,0 +1,122 @@ +{ + "model": "dememwm_sourcefree_countcorr_off_step200k", + "master_seed": 43, + "sample_seed_formula": "master_seed * 1000000 + global_test_index", + "test_clips": 300, + "source_offset": 100, + "context_range": [ + 100, + 700 + ], + "target_range": [ + 700, + 1200 + ], + "context_frames": 600, + "generated_frames": 500, + "local_window": 8, + "ddim_steps": 20, + "ddim_eta": 0.0, + "per_device_batch_size": 1, + "worker_ranges": [ + "0:10", + "10:20", + "20:30", + "30:40", + "40:50", + "50:60", + "60:70", + "70:80", + "80:90", + "90:100", + "100:110", + "110:120", + "120:129", + "129:138", + "138:147", + "147:156", + "156:165", + "165:174", + "174:183", + "183:192", + "192:201", + "201:210", + "210:219", + "219:228", + "228:237", + "237:246", + "246:255", + "255:264", + "264:273", + "273:282", + "282:291", + "291:300" + ], + "distribution": "32 disjoint standalone batch-1 workers; four per GPU; no DistributedSampler", + "generated_frame_bins": [ + [ + 0, + 100 + ], + [ + 100, + 200 + ], + [ + 200, + 300 + ], + [ + 300, + 400 + ], + [ + 400, + 500 + ] + ], + "reconstruction_metrics": { + "names": [ + "mse", + "psnr", + "lpips" + ], + "population_per_bin": { + "clips": 300, + "frames": 30000 + }, + "population_overall": { + "clips": 300, + "frames": 150000 + }, + "lpips_input_range": [ + 0.0, + 1.0 + ], + "lpips_normalize": true + }, + "frame_fid": { + "implementation": "TorchMetrics FrechetInceptionDistance", + "feature": 2048, + "normalize": false, + "input": "uint8 RGB [0,255]", + "official_compatible_bins": { + "clips": 50, + "frames_per_bin": 5000 + }, + "supplementary_pooled": { + "clips": 50, + "frames": 25000 + }, + "full_300_bins": { + "clips": 300, + "frames_per_bin": 30000 + }, + "full_300_pooled": { + "clips": 300, + "frames": 150000 + } + }, + "manifest_sha256": "ef4e4de9d3975d8a7e724bb609e57f20186cd1f7285457c00fb85c1aca973b99", + "fid50_manifest_sha256": "f9536911e434610cce23306ff99575db53cd7fc29e3df841da4b6631bf418634" +} diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..dae70b4071c2d2a834268c2a26f061d308f11f2c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.csv @@ -0,0 +1,301 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +50,50,212,test/val_savanna/1/000039.mp4,test/val_savanna/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +51,51,4,test/val_desert/0/000001.mp4,test/val_desert/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +52,52,7,test/val_desert/0/000002.mp4,test/val_desert/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +53,53,310,test/val_iceplains/0/000026.mp4,test/val_iceplains/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +54,54,284,test/val_iceplains/0/000015.mp4,test/val_iceplains/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +55,55,109,test/val_desert/3/000041.mp4,test/val_desert/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +56,56,45,test/val_desert/0/000018.mp4,test/val_desert/0/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +57,57,93,test/val_desert/0/000035.mp4,test/val_desert/0/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +58,58,400,test/val_plains/3/000011.mp4,test/val_plains/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +59,59,300,test/val_iceplains/2/000021.mp4,test/val_iceplains/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +60,60,226,test/val_savanna/2/000043.mp4,test/val_savanna/2/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +61,61,417,test/val_plains/2/000018.mp4,test/val_plains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +62,62,120,test/val_desert/3/000046.mp4,test/val_desert/3/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +63,63,106,test/val_desert/3/000040.mp4,test/val_desert/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +64,64,345,test/val_iceplains/1/000039.mp4,test/val_iceplains/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +65,65,169,test/val_savanna/2/000016.mp4,test/val_savanna/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +66,66,312,test/val_iceplains/1/000027.mp4,test/val_iceplains/1/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +67,67,368,test/val_iceplains/0/000048.mp4,test/val_iceplains/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +68,68,449,test/val_plains/1/000034.mp4,test/val_plains/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +69,69,11,test/val_desert/1/000004.mp4,test/val_desert/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +70,70,6,test/val_desert/2/000001.mp4,test/val_desert/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +71,71,257,test/val_iceplains/0/000005.mp4,test/val_iceplains/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +72,72,262,test/val_iceplains/3/000006.mp4,test/val_iceplains/3/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +73,73,164,test/val_savanna/0/000015.mp4,test/val_savanna/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +74,74,337,test/val_iceplains/2/000036.mp4,test/val_iceplains/2/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +75,75,95,test/val_desert/0/000036.mp4,test/val_desert/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +76,76,277,test/val_iceplains/0/000013.mp4,test/val_iceplains/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +77,77,26,test/val_desert/1/000010.mp4,test/val_desert/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +78,78,77,test/val_desert/0/000029.mp4,test/val_desert/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +79,79,294,test/val_iceplains/1/000019.mp4,test/val_iceplains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +80,80,34,test/val_desert/1/000014.mp4,test/val_desert/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +81,81,188,test/val_savanna/1/000029.mp4,test/val_savanna/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +82,82,170,test/val_savanna/0/000017.mp4,test/val_savanna/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +83,83,415,test/val_plains/2/000017.mp4,test/val_plains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +84,84,145,test/val_savanna/0/000006.mp4,test/val_savanna/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +85,85,133,test/val_savanna/1/000002.mp4,test/val_savanna/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +86,86,347,test/val_iceplains/2/000039.mp4,test/val_iceplains/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +87,87,209,test/val_savanna/0/000038.mp4,test/val_savanna/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +88,88,404,test/val_plains/1/000013.mp4,test/val_plains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +89,89,211,test/val_savanna/2/000038.mp4,test/val_savanna/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +90,90,464,test/val_plains/2/000044.mp4,test/val_plains/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +91,91,162,test/val_savanna/1/000014.mp4,test/val_savanna/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +92,92,419,test/val_plains/2/000019.mp4,test/val_plains/2/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +93,93,127,test/val_desert/3/000049.mp4,test/val_desert/3/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +94,94,370,test/val_iceplains/2/000048.mp4,test/val_iceplains/2/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +95,95,293,test/val_iceplains/2/000018.mp4,test/val_iceplains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +96,96,251,test/val_iceplains/1/000003.mp4,test/val_iceplains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +97,97,441,test/val_plains/1/000031.mp4,test/val_plains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +98,98,69,test/val_desert/2/000026.mp4,test/val_desert/2/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +99,99,286,test/val_iceplains/2/000015.mp4,test/val_iceplains/2/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +100,100,438,test/val_plains/0/000029.mp4,test/val_plains/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +101,101,376,test/val_plains/0/000002.mp4,test/val_plains/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +102,102,163,test/val_savanna/2/000014.mp4,test/val_savanna/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +103,103,31,test/val_desert/1/000013.mp4,test/val_desert/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +104,104,241,test/val_savanna/0/000049.mp4,test/val_savanna/0/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +105,105,458,test/val_plains/0/000040.mp4,test/val_plains/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +106,106,371,test/val_iceplains/1/000049.mp4,test/val_iceplains/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +107,107,230,test/val_savanna/3/000045.mp4,test/val_savanna/3/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +108,108,372,test/val_iceplains/2/000049.mp4,test/val_iceplains/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +109,109,223,test/val_savanna/1/000042.mp4,test/val_savanna/1/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +110,110,68,test/val_desert/0/000026.mp4,test/val_desert/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +111,111,461,test/val_plains/2/000042.mp4,test/val_plains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +112,112,151,test/val_savanna/1/000009.mp4,test/val_savanna/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +113,113,72,test/val_desert/3/000027.mp4,test/val_desert/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +114,114,427,test/val_plains/0/000024.mp4,test/val_plains/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +115,115,317,test/val_iceplains/3/000029.mp4,test/val_iceplains/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +116,116,249,test/val_iceplains/2/000001.mp4,test/val_iceplains/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +117,117,364,test/val_iceplains/2/000046.mp4,test/val_iceplains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +118,118,96,test/val_desert/2/000037.mp4,test/val_desert/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +119,119,190,test/val_savanna/3/000029.mp4,test/val_savanna/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +120,120,227,test/val_savanna/2/000044.mp4,test/val_savanna/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +121,121,244,test/val_iceplains/0/000000.mp4,test/val_iceplains/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +122,122,38,test/val_desert/3/000015.mp4,test/val_desert/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +123,123,291,test/val_iceplains/2/000017.mp4,test/val_iceplains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +124,124,85,test/val_desert/2/000032.mp4,test/val_desert/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +125,125,358,test/val_iceplains/0/000044.mp4,test/val_iceplains/0/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +126,126,50,test/val_desert/1/000020.mp4,test/val_desert/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +127,127,207,test/val_savanna/0/000036.mp4,test/val_savanna/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +128,128,330,test/val_iceplains/2/000034.mp4,test/val_iceplains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +129,129,132,test/val_savanna/0/000001.mp4,test/val_savanna/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +130,130,423,test/val_plains/3/000021.mp4,test/val_plains/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +131,131,66,test/val_desert/2/000025.mp4,test/val_desert/2/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +132,132,153,test/val_savanna/3/000009.mp4,test/val_savanna/3/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +133,133,382,test/val_plains/0/000004.mp4,test/val_plains/0/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +134,134,143,test/val_savanna/2/000005.mp4,test/val_savanna/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +135,135,348,test/val_iceplains/1/000040.mp4,test/val_iceplains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +136,136,384,test/val_plains/2/000005.mp4,test/val_plains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +137,137,30,test/val_desert/0/000012.mp4,test/val_desert/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +138,138,203,test/val_savanna/2/000034.mp4,test/val_savanna/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +139,139,179,test/val_savanna/0/000022.mp4,test/val_savanna/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +140,140,253,test/val_iceplains/2/000003.mp4,test/val_iceplains/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +141,141,289,test/val_iceplains/2/000016.mp4,test/val_iceplains/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +142,142,422,test/val_plains/2/000020.mp4,test/val_plains/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +143,143,20,test/val_desert/0/000008.mp4,test/val_desert/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +144,144,130,test/val_savanna/0/000000.mp4,test/val_savanna/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +145,145,296,test/val_iceplains/1/000020.mp4,test/val_iceplains/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +146,146,278,test/val_iceplains/3/000013.mp4,test/val_iceplains/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +147,147,269,test/val_iceplains/0/000009.mp4,test/val_iceplains/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +148,148,314,test/val_iceplains/0/000028.mp4,test/val_iceplains/0/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +149,149,281,test/val_iceplains/0/000014.mp4,test/val_iceplains/0/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +150,150,149,test/val_savanna/0/000008.mp4,test/val_savanna/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +151,151,184,test/val_savanna/0/000024.mp4,test/val_savanna/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +152,152,407,test/val_plains/1/000014.mp4,test/val_plains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +153,153,304,test/val_iceplains/2/000023.mp4,test/val_iceplains/2/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +154,154,208,test/val_savanna/3/000037.mp4,test/val_savanna/3/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +155,155,182,test/val_savanna/3/000023.mp4,test/val_savanna/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +156,156,394,test/val_plains/1/000009.mp4,test/val_plains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +157,157,14,test/val_desert/0/000006.mp4,test/val_desert/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +158,158,55,test/val_desert/3/000021.mp4,test/val_desert/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +159,159,75,test/val_desert/2/000028.mp4,test/val_desert/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +160,160,319,test/val_iceplains/1/000031.mp4,test/val_iceplains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +161,161,39,test/val_desert/1/000016.mp4,test/val_desert/1/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +162,162,418,test/val_plains/1/000019.mp4,test/val_plains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +163,163,389,test/val_plains/3/000007.mp4,test/val_plains/3/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +164,164,409,test/val_plains/2/000014.mp4,test/val_plains/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +165,165,344,test/val_iceplains/2/000038.mp4,test/val_iceplains/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +166,166,396,test/val_plains/2/000009.mp4,test/val_plains/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +167,167,67,test/val_desert/1/000026.mp4,test/val_desert/1/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +168,168,99,test/val_desert/3/000038.mp4,test/val_desert/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +169,169,288,test/val_iceplains/3/000016.mp4,test/val_iceplains/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +170,170,97,test/val_desert/1/000038.mp4,test/val_desert/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +171,171,178,test/val_savanna/2/000021.mp4,test/val_savanna/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +172,172,264,test/val_iceplains/1/000007.mp4,test/val_iceplains/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +173,173,266,test/val_iceplains/0/000008.mp4,test/val_iceplains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +174,174,303,test/val_iceplains/1/000023.mp4,test/val_iceplains/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +175,175,442,test/val_plains/0/000031.mp4,test/val_plains/0/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +176,176,233,test/val_savanna/1/000047.mp4,test/val_savanna/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +177,177,240,test/val_savanna/1/000049.mp4,test/val_savanna/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +178,178,42,test/val_desert/0/000017.mp4,test/val_desert/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +179,179,369,test/val_iceplains/3/000048.mp4,test/val_iceplains/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +180,180,398,test/val_plains/3/000010.mp4,test/val_plains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +181,181,435,test/val_plains/2/000027.mp4,test/val_plains/2/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +182,182,242,test/val_savanna/2/000049.mp4,test/val_savanna/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +183,183,361,test/val_iceplains/2/000045.mp4,test/val_iceplains/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +184,184,470,test/val_plains/2/000047.mp4,test/val_plains/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +185,185,356,test/val_iceplains/2/000042.mp4,test/val_iceplains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +186,186,468,test/val_plains/2/000046.mp4,test/val_plains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +187,187,113,test/val_desert/0/000043.mp4,test/val_desert/0/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +188,188,19,test/val_desert/1/000008.mp4,test/val_desert/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +189,189,199,test/val_savanna/0/000033.mp4,test/val_savanna/0/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +190,190,118,test/val_desert/2/000045.mp4,test/val_desert/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +191,191,433,test/val_plains/0/000027.mp4,test/val_plains/0/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +192,192,328,test/val_iceplains/0/000034.mp4,test/val_iceplains/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +193,193,213,test/val_savanna/0/000039.mp4,test/val_savanna/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +194,194,339,test/val_iceplains/0/000037.mp4,test/val_iceplains/0/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +195,195,466,test/val_plains/1/000046.mp4,test/val_plains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +196,196,355,test/val_iceplains/3/000042.mp4,test/val_iceplains/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +197,197,336,test/val_iceplains/3/000036.mp4,test/val_iceplains/3/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +198,198,420,test/val_plains/0/000020.mp4,test/val_plains/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +199,199,410,test/val_plains/1/000015.mp4,test/val_plains/1/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +200,200,27,test/val_desert/1/000011.mp4,test/val_desert/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +201,201,86,test/val_desert/1/000033.mp4,test/val_desert/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +202,202,210,test/val_savanna/3/000038.mp4,test/val_savanna/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +203,203,103,test/val_desert/3/000039.mp4,test/val_desert/3/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +204,204,245,test/val_iceplains/3/000000.mp4,test/val_iceplains/3/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +205,205,25,test/val_desert/2/000009.mp4,test/val_desert/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +206,206,340,test/val_iceplains/2/000037.mp4,test/val_iceplains/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +207,207,167,test/val_savanna/0/000016.mp4,test/val_savanna/0/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +208,208,424,test/val_plains/0/000022.mp4,test/val_plains/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +209,209,448,test/val_plains/2/000033.mp4,test/val_plains/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +210,210,386,test/val_plains/0/000006.mp4,test/val_plains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +211,211,392,test/val_plains/3/000008.mp4,test/val_plains/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +212,212,105,test/val_desert/1/000040.mp4,test/val_desert/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +213,213,267,test/val_iceplains/2/000008.mp4,test/val_iceplains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +214,214,90,test/val_desert/0/000034.mp4,test/val_desert/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +215,215,426,test/val_plains/0/000023.mp4,test/val_plains/0/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +216,216,168,test/val_savanna/3/000016.mp4,test/val_savanna/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +217,217,218,test/val_savanna/3/000040.mp4,test/val_savanna/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +218,218,445,test/val_plains/0/000032.mp4,test/val_plains/0/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +219,219,48,test/val_desert/3/000019.mp4,test/val_desert/3/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +220,220,104,test/val_desert/2/000039.mp4,test/val_desert/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +221,221,206,test/val_savanna/2/000035.mp4,test/val_savanna/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +222,222,228,test/val_savanna/1/000045.mp4,test/val_savanna/1/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +223,223,200,test/val_savanna/2/000033.mp4,test/val_savanna/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +224,224,62,test/val_desert/2/000024.mp4,test/val_desert/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +225,225,121,test/val_desert/1/000047.mp4,test/val_desert/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +226,226,147,test/val_savanna/0/000007.mp4,test/val_savanna/0/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +227,227,146,test/val_savanna/2/000006.mp4,test/val_savanna/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +228,228,225,test/val_savanna/1/000043.mp4,test/val_savanna/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +229,229,313,test/val_iceplains/3/000027.mp4,test/val_iceplains/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +230,230,354,test/val_iceplains/0/000042.mp4,test/val_iceplains/0/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +231,231,367,test/val_iceplains/1/000048.mp4,test/val_iceplains/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +232,232,297,test/val_iceplains/3/000020.mp4,test/val_iceplains/3/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +233,233,124,test/val_desert/1/000048.mp4,test/val_desert/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +234,234,171,test/val_savanna/2/000017.mp4,test/val_savanna/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +235,235,342,test/val_iceplains/0/000038.mp4,test/val_iceplains/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +236,236,60,test/val_desert/3/000023.mp4,test/val_desert/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +237,237,131,test/val_savanna/1/000001.mp4,test/val_savanna/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +238,238,365,test/val_iceplains/1/000047.mp4,test/val_iceplains/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +239,239,315,test/val_iceplains/2/000028.mp4,test/val_iceplains/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +240,240,220,test/val_savanna/1/000041.mp4,test/val_savanna/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +241,241,8,test/val_desert/1/000003.mp4,test/val_desert/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +242,242,261,test/val_iceplains/0/000006.mp4,test/val_iceplains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +243,243,17,test/val_desert/1/000007.mp4,test/val_desert/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +244,244,462,test/val_plains/1/000043.mp4,test/val_plains/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +245,245,454,test/val_plains/3/000038.mp4,test/val_plains/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +246,246,351,test/val_iceplains/1/000041.mp4,test/val_iceplains/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +247,247,383,test/val_plains/3/000005.mp4,test/val_plains/3/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +248,248,18,test/val_desert/2/000007.mp4,test/val_desert/2/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +249,249,457,test/val_plains/1/000040.mp4,test/val_plains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +250,250,326,test/val_iceplains/3/000033.mp4,test/val_iceplains/3/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +251,251,254,test/val_iceplains/1/000004.mp4,test/val_iceplains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +252,252,375,test/val_plains/1/000002.mp4,test/val_plains/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +253,253,141,test/val_savanna/0/000005.mp4,test/val_savanna/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +254,254,306,test/val_iceplains/2/000024.mp4,test/val_iceplains/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +255,255,10,test/val_desert/2/000003.mp4,test/val_desert/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +256,256,173,test/val_savanna/0/000020.mp4,test/val_savanna/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +257,257,320,test/val_iceplains/1/000032.mp4,test/val_iceplains/1/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +258,258,115,test/val_desert/1/000044.mp4,test/val_desert/1/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +259,259,341,test/val_iceplains/1/000038.mp4,test/val_iceplains/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +260,260,61,test/val_desert/3/000024.mp4,test/val_desert/3/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +261,261,231,test/val_savanna/1/000046.mp4,test/val_savanna/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +262,262,224,test/val_savanna/3/000042.mp4,test/val_savanna/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +263,263,292,test/val_iceplains/3/000018.mp4,test/val_iceplains/3/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +264,264,334,test/val_iceplains/1/000036.mp4,test/val_iceplains/1/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +265,265,159,test/val_savanna/0/000013.mp4,test/val_savanna/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +266,266,102,test/val_desert/0/000039.mp4,test/val_desert/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +267,267,362,test/val_iceplains/1/000046.mp4,test/val_iceplains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +268,268,322,test/val_iceplains/3/000032.mp4,test/val_iceplains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +269,269,381,test/val_plains/1/000004.mp4,test/val_plains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +270,270,84,test/val_desert/3/000032.mp4,test/val_desert/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +271,271,243,test/val_iceplains/1/000000.mp4,test/val_iceplains/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +272,272,65,test/val_desert/3/000025.mp4,test/val_desert/3/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +273,273,64,test/val_desert/0/000025.mp4,test/val_desert/0/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +274,274,272,test/val_iceplains/2/000010.mp4,test/val_iceplains/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +275,275,399,test/val_plains/0/000011.mp4,test/val_plains/0/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +276,276,248,test/val_iceplains/3/000001.mp4,test/val_iceplains/3/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +277,277,43,test/val_desert/3/000017.mp4,test/val_desert/3/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +278,278,393,test/val_plains/2/000008.mp4,test/val_plains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +279,279,123,test/val_desert/2/000047.mp4,test/val_desert/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +280,280,29,test/val_desert/2/000011.mp4,test/val_desert/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +281,281,122,test/val_desert/0/000047.mp4,test/val_desert/0/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +282,282,157,test/val_savanna/3/000011.mp4,test/val_savanna/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +283,283,333,test/val_iceplains/2/000035.mp4,test/val_iceplains/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +284,284,307,test/val_iceplains/1/000025.mp4,test/val_iceplains/1/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +285,285,9,test/val_desert/0/000003.mp4,test/val_desert/0/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +286,286,250,test/val_iceplains/2/000002.mp4,test/val_iceplains/2/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +287,287,165,test/val_savanna/3/000015.mp4,test/val_savanna/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +288,288,0,test/val_desert/1/000000.mp4,test/val_desert/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +289,289,271,test/val_iceplains/3/000010.mp4,test/val_iceplains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +290,290,276,test/val_iceplains/1/000013.mp4,test/val_iceplains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +291,291,82,test/val_desert/1/000031.mp4,test/val_desert/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +292,292,191,test/val_savanna/2/000029.mp4,test/val_savanna/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +293,293,374,test/val_plains/1/000001.mp4,test/val_plains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +294,294,446,test/val_plains/3/000032.mp4,test/val_plains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +295,295,378,test/val_plains/1/000003.mp4,test/val_plains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +296,296,152,test/val_savanna/0/000009.mp4,test/val_savanna/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +297,297,54,test/val_desert/0/000021.mp4,test/val_desert/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +298,298,259,test/val_iceplains/2/000005.mp4,test/val_iceplains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +299,299,91,test/val_desert/2/000034.mp4,test/val_desert/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.sha256 b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..065350069e5b0f46e33cd51ad63f60aa956822fd --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.sha256 @@ -0,0 +1 @@ +ef4e4de9d3975d8a7e724bb609e57f20186cd1f7285457c00fb85c1aca973b99 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_0.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_0.log new file mode 100644 index 0000000000000000000000000000000000000000..5ba222a3e2a2c0ece2e200c43cfe680b535eac00 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_0.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:08,007][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:16,418][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:16,563][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:16,563][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:16,563][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:16,563][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:20,517][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=0', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=0', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_1.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_1.log new file mode 100644 index 0000000000000000000000000000000000000000..af76110435df64ea35e730b2c381f3c6b60d5bdf --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_1.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:09,229][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:17,789][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:17,934][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:17,934][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:17,934][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:17,934][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:21,148][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=10', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=10', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_10.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_10.log new file mode 100644 index 0000000000000000000000000000000000000000..0824b91823dde0d7b91b42083a000b2bcff0cf55 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_10.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:00,171][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:08,633][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:08,773][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:08,773][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:08,773][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:08,773][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:14,151][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=100', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=100', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_11.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_11.log new file mode 100644 index 0000000000000000000000000000000000000000..cf567f4a5a78922444ef6bbb89efc8ebc399b86d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_11.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:03:59,441][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:07,726][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:07,861][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:07,861][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:07,861][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:07,861][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:12,968][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=110', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=110', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_12.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_12.log new file mode 100644 index 0000000000000000000000000000000000000000..7583d860c18560a00bb63e0b650dd98b641aadb8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_12.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:06,051][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:14,446][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:14,584][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:14,584][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:14,584][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:14,584][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:18,320][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=120', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=120', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_13.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_13.log new file mode 100644 index 0000000000000000000000000000000000000000..17155040d4b13536cb239442a6e72da587b1f8c3 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_13.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:26,403][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:34,730][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:34,873][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:34,873][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:34,873][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:34,873][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:38,105][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=129', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=129', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_14.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_14.log new file mode 100644 index 0000000000000000000000000000000000000000..5c94fb1744d2644978027044f13b581001775963 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_14.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:17,367][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:25,831][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:25,968][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:25,968][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:25,968][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:25,968][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:28,861][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=138', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=138', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_15.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_15.log new file mode 100644 index 0000000000000000000000000000000000000000..8ca6c41414ac0d790eb48171aed72b4fb2faae8f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_15.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:12,880][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:21,257][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:21,396][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:21,396][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:21,396][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:21,396][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:25,213][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=147', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=147', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_16.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_16.log new file mode 100644 index 0000000000000000000000000000000000000000..acb951d84d007d6c386b587726372a4af3dd1423 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_16.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:08,432][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:16,903][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:17,059][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:17,059][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:17,059][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:17,059][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:20,528][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=156', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=156', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_17.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_17.log new file mode 100644 index 0000000000000000000000000000000000000000..898fd49b8bd656bc14c779afcab89a7bd1c0e117 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_17.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:04,444][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:12,901][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:13,041][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:13,041][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:13,041][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:13,041][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:16,300][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=165', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=165', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_18.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_18.log new file mode 100644 index 0000000000000000000000000000000000000000..5a4f19f90b1d97dbd77769ced4fe5f927dc90b84 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_18.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:01,283][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:09,595][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:09,739][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:09,739][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:09,739][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:09,739][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:14,760][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=174', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=174', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_19.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_19.log new file mode 100644 index 0000000000000000000000000000000000000000..4919720e58587323a841c016b376dcf6087572bd --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_19.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:06,548][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:14,892][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:15,035][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:15,035][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:15,035][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:15,035][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:18,565][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=183', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=183', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_2.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_2.log new file mode 100644 index 0000000000000000000000000000000000000000..67aab15e1430c182b0a4bac9b7e18f57a857dbee --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_2.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:10,347][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:18,686][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:18,828][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:18,828][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:18,828][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:18,828][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:21,726][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=20', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=20', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_20.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_20.log new file mode 100644 index 0000000000000000000000000000000000000000..a28da9e6d9844881a64f5a06f9678526d839a38b --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_20.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:03:59,095][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:07,428][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:07,564][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:07,564][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:07,565][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:07,565][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:12,277][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=192', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=192', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_21.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_21.log new file mode 100644 index 0000000000000000000000000000000000000000..3d16420fbda306b38f2a5d8e9f6107ed4fea5aa8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_21.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:19,850][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:28,200][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:28,352][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:28,352][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:28,352][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:28,352][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:31,319][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=201', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=201', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_22.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_22.log new file mode 100644 index 0000000000000000000000000000000000000000..c6936b6197d870648504bd1bdec86669a9b84743 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_22.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:14,670][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:23,090][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:23,221][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:23,221][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:23,221][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:23,221][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:26,371][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=210', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=210', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_23.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_23.log new file mode 100644 index 0000000000000000000000000000000000000000..ed782359cbd9ee0de038cb5b92bc0f608353be47 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_23.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:06,821][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:15,229][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:15,366][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:15,366][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:15,366][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:15,366][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:19,304][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=219', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=219', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_24.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_24.log new file mode 100644 index 0000000000000000000000000000000000000000..ecb629b378138f869abad9fcf8609ff2e1d10eff --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_24.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:04,717][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:13,032][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:13,168][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:13,169][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:13,169][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:13,169][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:16,529][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=228', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=228', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_25.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_25.log new file mode 100644 index 0000000000000000000000000000000000000000..06cb94f4b89e8dae79ecbf171f37c0985ca42ea6 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_25.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:25,316][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:33,771][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:33,907][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:33,908][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:33,908][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:33,908][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:37,128][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=237', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=237', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_26.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_26.log new file mode 100644 index 0000000000000000000000000000000000000000..4b8c60771d4bb232b9c0ed631f1d109ef6152351 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_26.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:01,050][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:09,350][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:09,496][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:09,497][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:09,497][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:09,497][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:14,177][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=246', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=246', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_27.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_27.log new file mode 100644 index 0000000000000000000000000000000000000000..813252166c1f609d9fa4f14533e3c38124520d78 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_27.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:03:59,243][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:07,561][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:07,701][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:07,701][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:07,701][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:07,701][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:13,538][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=255', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=255', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_28.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_28.log new file mode 100644 index 0000000000000000000000000000000000000000..a3eb4b9cd8602c76c31a268e85dfe2d067181e5a --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_28.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:12,303][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:20,781][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:20,914][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:20,914][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:20,914][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:20,914][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:24,458][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=264', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=264', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_29.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_29.log new file mode 100644 index 0000000000000000000000000000000000000000..1d42d47fb9f77de625786287d1f8db3791a44dd8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_29.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:03:59,371][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:07,881][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:08,049][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:08,049][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:08,049][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:08,049][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:14,110][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=273', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=273', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_3.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_3.log new file mode 100644 index 0000000000000000000000000000000000000000..62feb0d6d2b48c569c3ea0d8c5810316fd812ea8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_3.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:15,995][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:24,392][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:24,535][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:24,535][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:24,535][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:24,535][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:27,842][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=30', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=30', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_30.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_30.log new file mode 100644 index 0000000000000000000000000000000000000000..cf21bc1a6970f7e24816eab152d594dedf1ef87f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_30.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:23,981][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:32,345][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:32,489][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:32,490][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:32,490][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:32,490][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:35,654][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=282', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=282', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_31.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_31.log new file mode 100644 index 0000000000000000000000000000000000000000..3126702411c455bdd18547beac1ec8ce90123e9a --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_31.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:13,226][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:21,648][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:21,798][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:21,798][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:21,798][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:21,798][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:25,339][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=291', '+algorithm.eval_manifest_count=9', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=291', '+experiment.test.manifest_count=9', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=9', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_4.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_4.log new file mode 100644 index 0000000000000000000000000000000000000000..dfaa6024e0639b7497ddf42b52d67ecba008946a --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_4.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:20,670][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:29,061][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:29,204][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:29,204][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:29,204][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:29,204][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:32,175][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=40', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=40', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_5.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_5.log new file mode 100644 index 0000000000000000000000000000000000000000..2dfd007687da8c9c7c94a050e642fbe14251555f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_5.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:22,050][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:30,445][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:30,595][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:30,595][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:30,596][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:30,596][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:33,663][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=50', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=50', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_6.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_6.log new file mode 100644 index 0000000000000000000000000000000000000000..3ecb0735bdfcf24f789ce084bfa20399cdfaba53 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_6.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:18,656][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:27,097][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:27,244][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:27,244][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:27,244][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:27,244][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:30,813][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=60', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=60', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_7.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_7.log new file mode 100644 index 0000000000000000000000000000000000000000..b6cb0f602a39695eec5dae8a5a808878ffee2b9a --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_7.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:18,932][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:27,330][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:27,475][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:27,475][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:27,475][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:27,475][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:30,832][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=70', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=70', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_8.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_8.log new file mode 100644 index 0000000000000000000000000000000000000000..2ee3ae814e2940cd1d108cb996688f0e8feff03d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_8.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:03,444][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:12,292][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:12,456][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:12,456][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:12,456][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:12,456][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:15,488][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=80', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=80', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_9.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_9.log new file mode 100644 index 0000000000000000000000000000000000000000..d09bf3030db124dff6f58e5d92aa27a8b01e08fd --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/worker_9.log @@ -0,0 +1,64 @@ +INFO: [rank: 0] Seed set to 43 +[2026-07-22 14:04:02,693][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +Outputs will be saved to: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09/hydra/run +Will load checkpoint from /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Executing task: test out of ['test'] +[2026-07-22 14:04:10,967][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:11,125][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:11,125][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:11,125][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:11,125][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:15,048][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +Error executing job with overrides: ['+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43', 'load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt', 'auto_resume=false', '+customized_load=false', 'wandb.mode=disabled', 'experiment.tasks=[test]', 'experiment.num_nodes=1', '+seed=43', 'dataset=video_minecraft_dememwm_latent', 'dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft', 'dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features', 'dataset.n_frames=8', 'dataset.n_frames_valid=1100', 'dataset.context_length=100', 'dataset.single_eval_clip=true', 'dataset.memory_selection.enabled=true', 'dataset.memory_selection.causal=true', 'dataset.memory_selection.max_anchor_frames=2', 'dataset.memory_selection.max_dynamic_frames=4', 'dataset.memory_selection.max_revisit_frames=2', 'dataset.memory_selection.training_dropout=0.1', 'dataset.memory_selection.dynamic.selection_policy=multiview', 'dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse', '+dataset.seed=42', '+dataset.wo_updown=false', 'algorithm=dememwm_base', 'algorithm.causal=true', 'algorithm.n_tokens=8', 'algorithm.context_frames=600', 'algorithm.metric_report_segment=100', 'algorithm.use_memory_attention=true', 'algorithm.anchor_zero_geometry=false', 'algorithm.memory_role_embedding=true', 'algorithm.memory_source_embedding=false', 'algorithm.memory_attention_key_only_geometry=true', 'algorithm.memory_age.enabled=true', 'algorithm.memory_age.num_buckets=32', 'algorithm.memory_age.num_exact_buckets=17', 'algorithm.memory_age.overflow_start=100', 'algorithm.memory_stream_count_correction=false', 'algorithm.noise_route.anchor=high', 'algorithm.noise_route.dynamic=low', 'algorithm.noise_route.revisit=all', 'algorithm.memory_source_simulation.training_generated_suffix_max_frames=null', 'algorithm.memory_noise.enabled=true', 'algorithm.memory_noise.mode=source_free_logsnr', 'algorithm.memory_noise.stable_probability=0.25', 'algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1', 'algorithm.memory_noise.validation_noisy_memory=false', 'algorithm.chunk_size=1', 'algorithm.diffusion.sampling_timesteps=20', 'algorithm.diffusion.ddim_sampling_eta=0.0', 'algorithm.log_video=false', 'algorithm.save_local=true', '+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09', '+algorithm.lpips_batch_size=16', '+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+algorithm.eval_manifest_start=90', '+algorithm.eval_manifest_count=10', '+algorithm.eval_master_seed=43', '+algorithm.eval_batch_size=1', '+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv', '+experiment.test.manifest_start=90', '+experiment.test.manifest_count=10', 'experiment.test.batch_size=1', 'experiment.test.limit_batch=10', 'experiment.test.data.num_workers=2', 'experiment.test.inference_mode=false'] +Traceback (most recent call last): + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 338, in + run() # pylint: disable=no-value-for-parameter + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/main.py", line 94, in decorated_main + _run_hydra( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra + _run_app( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 457, in _run_app + run_and_report( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 223, in run_and_report + raise ex + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 220, in run_and_report + return func() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/utils.py", line 458, in + lambda: hydra.run( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run + _ = ret.return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value + raise self._return_value + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job + ret.return_value = task_function(task_cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 334, in run + run_local(cfg) + File "/proj/cvl/users/x_fahkh2/DeMemWM/main.py", line 258, in run_local + experiment.exec_task(task) + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 593, in exec_task + getattr(self, task)() + File "/proj/cvl/users/x_fahkh2/DeMemWM/experiments/exp_base.py", line 973, in test + trainer.test( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 754, in test + return call._call_and_handle_interrupt( + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 44, in _call_and_handle_interrupt + return trainer_fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 794, in _test_impl + results = self._run(model, ckpt_path=ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/trainer.py", line 955, in _run + self._checkpoint_connector._restore_modules_and_callbacks(ckpt_path) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 396, in _restore_modules_and_callbacks + self.restore_model() + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/checkpoint_connector.py", line 273, in restore_model + call._call_lightning_module_hook(trainer, "on_load_checkpoint", self._loaded_checkpoint) + File "/proj/cvl/users/x_fahkh2/envs/worldmem/lib/python3.10/site-packages/lightning/pytorch/trainer/call.py", line 157, in _call_lightning_module_hook + output = fn(*args, **kwargs) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 1468, in on_load_checkpoint + _validate_dememwm_training_contract(self.cfg, checkpoint) + File "/proj/cvl/users/x_fahkh2/DeMemWM/algorithms/dememwm/df_video.py", line 376, in _validate_dememwm_training_contract + raise RuntimeError(f"DeMemWM training contract mismatch: {mismatch}") +RuntimeError: DeMemWM training contract mismatch: dememwm_training_contract.trainability.full_dit_start_step: expected NoneType, found int diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e5d93a8e7c185795cbd97143ab64b62a31bc6972 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 0 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 0 + eval_manifest_count: 10 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fd27ea7d8856c7c920014cea0b2fd7d87dcca06a --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=0 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=0 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=0,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=0,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5d68dcad54f6ade1e6b9c6941dfc4a7ff6b2f4d6 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=0 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=0 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..284dea1b9b62692f7c52b3f4ea1380ab810930f6 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_00/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:08,007][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:16,418][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:16,563][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:16,563][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:16,563][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:16,563][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:20,517][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c3681045a16688953172e6ca3442dc95f3e85f1d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 10 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 10 + eval_manifest_count: 10 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d49bf1ccfcb5b9402547489a8a4e9a7bb4ce1990 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=10 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=10 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=10,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=10,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4e082853bb4338efa726ef2d4f426c3b3fb50cff --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=10 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=10 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..9307a96a5461264f32fb4dfc5401730ddd8ceb5f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_01/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:09,229][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:17,789][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:17,934][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:17,934][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:17,934][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:17,934][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:21,148][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4e77e280324931b9645bab1cd7ccd2c6d9562aa4 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 20 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 20 + eval_manifest_count: 10 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..79f9e1016531148817837325a4de947352b406d2 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=20 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=20 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=20,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=20,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8001c2238e69bea0e6545d701aa49ec7bf25d1be --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=20 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=20 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..35bd76f96712a3793a3debbdba5be7cc2e42f3ce --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_02/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:10,347][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:18,686][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:18,828][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:18,828][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:18,828][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:18,828][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:21,726][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..602dfa2e5a9776404fc58c5824c9ab86dec5c111 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 30 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 30 + eval_manifest_count: 10 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..aaa5cb03f21ea16d54737f8d09b6d840df6a19a2 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=30 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=30 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=30,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=30,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cd37cd620e60a8df55a65ed7fa505fdea8342d70 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=30 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=30 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..a18245dd4ac95fc2b4831b727fb9e2a2147731fe --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_03/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:15,995][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:24,392][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:24,535][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:24,535][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:24,535][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:24,535][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:27,842][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4c16556f5ea4b64b59381d32ebf85b5f13a0f80b --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 40 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 40 + eval_manifest_count: 10 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9af3602b0bee8230bf777cff07d0da36be05c66d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=40 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=40 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=40,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=40,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c8895731404934a9d81263cc491c38974816400a --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=40 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=40 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..6f53c9c7a413a880628f1aadeda8a722164e9342 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_04/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:20,670][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:29,061][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:29,204][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:29,204][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:29,204][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:29,204][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:32,175][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8c0988da84358a8fd34dcaf790457792e436317c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 50 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 50 + eval_manifest_count: 10 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3823d8980e6865c77e4303f17a52e6030cb770ec --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=50 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=50 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=50,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=50,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..2470c5b16052889252b81599669ecc882b4bbbf2 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=50 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=50 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..cb4c0e439cef472b8a457aa8fe28ba5b801ea3b1 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_05/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:22,050][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:30,445][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:30,595][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:30,595][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:30,596][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:30,596][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:33,663][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e779cd75b99bfdb85d2146d0e133e16d9edb6def --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 60 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 60 + eval_manifest_count: 10 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c426bfad208ccf7cb59c09159c972a41f6e5bdfa --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=60 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=60 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=60,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=60,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e5d82b2851884fd26bacf9f1f5a1d13d21295cc8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=60 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=60 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..fa17fff0327981e1d3d77ffee3e28126b1cf6475 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_06/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:18,656][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:27,097][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:27,244][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:27,244][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:27,244][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:27,244][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:30,813][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..75850302f857893959e00c4bfc2482f887915b8f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 70 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 70 + eval_manifest_count: 10 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..750285d5ff9f44cfd74ab0624400d0b9523cf5bd --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=70 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=70 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=70,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=70,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8e00e5ee2e9c03eabff7e0f55a5b7f10e45c4a1f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=70 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=70 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..2fdc71e5f3ef615e2f72751b08d5fa731fa90e43 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_07/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:18,932][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:27,330][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:27,475][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:27,475][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:27,475][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:27,475][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:30,832][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..bd7336214446a161e6483a04f97a7bd10d0e077f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 80 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 80 + eval_manifest_count: 10 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0531bd6d3ab0201e7f2e807f6d2a642f18ba762b --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=80 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=80 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=80,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=80,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..92c3ba269b922bfdca35320110b4c1b22620e9c7 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=80 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=80 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..f9c153f130f200b6399240335a675f34a5909e01 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_08/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:03,444][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:12,292][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:12,456][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:12,456][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:12,456][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:12,456][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:15,488][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..49bd23a07a8e11aef9b67a9c31b5d26e8bff84ab --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 90 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 90 + eval_manifest_count: 10 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fcb9a8feb3ffbfc6b1679fb94f8e3979c514a942 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=90 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=90 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=90,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=90,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..28a9b2646ed2a2a2192d674f08d1b1078f71c75b --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=90 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=90 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..3bee6b8c99f87a262990ffdebdd4c1aa1e53115b --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_09/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:02,693][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:10,967][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:11,125][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:11,125][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:11,125][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:11,125][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:15,048][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0299babdc11ee16aa5a61f3c8f7bef6320544f8d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 100 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 100 + eval_manifest_count: 10 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ecad90e47fb440a9ab59d664369e4782481fb414 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=100 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=100 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=100,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=100,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..319084b537186ab5577d243e814ae4f5585edf4c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=100 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=100 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..bfea81c26e7e90e9b311f11379cb5c1fab2cb109 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_10/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:00,171][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:08,633][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:08,773][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:08,773][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:08,773][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:08,773][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:14,151][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..2a5a9246b0b7a0f14060655782e2712111d1ede1 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 10 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 110 + manifest_count: 10 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 110 + eval_manifest_count: 10 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0964cacf76f9fd46917a856eeb76e09680184289 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=110 + - +algorithm.eval_manifest_count=10 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=110 + - +experiment.test.manifest_count=10 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=10 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=10,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=110,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=10,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=110,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=10,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..1cbe2d201391480d96600043a4893ce5751ff11a --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=110 +- +algorithm.eval_manifest_count=10 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=110 +- +experiment.test.manifest_count=10 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=10 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..c9574b15a4e789580111b1cf8be16410dee531cd --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_11/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:03:59,441][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:07,726][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:07,861][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:07,861][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:07,861][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:07,861][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:12,968][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..52eb96678d097e35c8615f026dd36f738495c608 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 120 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 120 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..73fe254e8088a6616c4da0cad7e1e680f4bee5b0 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=120 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=120 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=120,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=120,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c9d0d29ffdcf28cdd7abb64cd31479cd3843ffb5 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=120 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=120 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..8a7a98ebcf070656aaf51f444d60cd4e30afdc00 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_12/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:06,051][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:14,446][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:14,584][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:14,584][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:14,584][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:14,584][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:18,320][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e57235b4a9a19580e82c5524071a25c56effc815 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 129 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 129 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..68d3746c9817d00bbe7ec206bd2e010a5f83c9d4 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=129 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=129 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=129,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=129,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..69e37c94f26909b5fefb80aa37662f19a7587dac --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=129 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=129 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..f79a76a36c75c82f44697689d9158da47e807889 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_13/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:26,403][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:34,730][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:34,873][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:34,873][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:34,873][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:34,873][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:38,105][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d806969e23a3da61682a4fb8a1245659666c489d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 138 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 138 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7985347f5a80d94fae15a9ec4a5498c0feaef37e --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=138 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=138 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=138,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=138,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f28d8bcb404ea9807d893083e07bb5b660dc701e --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=138 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=138 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..7fcb433b9bbdcb54a0c9112c0d8b375083c3833f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_14/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:17,367][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:25,831][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:25,968][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:25,968][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:25,968][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:25,968][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:28,861][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7085eb553628633bff11b60742adc7adee566f18 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 147 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 147 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c73805cac37a28176f7ad8b6d29c6b17b2f277cb --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=147 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=147 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=147,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=147,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..69dcb930e083973e9a7c530b7442a50ee5f94c24 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=147 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=147 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..1902c1910a6462e2bea97be03943c3c0f4150a30 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_15/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:12,880][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:21,257][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:21,396][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:21,396][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:21,396][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:21,396][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:25,213][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c9853d09d7e05bddd407fa303364ac11cc4c70f8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 156 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 156 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9a8247a36af4c4eeba0e15e1399d58fd6eb113f4 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=156 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=156 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=156,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=156,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ae957b6ad1f82531e1350cbc36d536238a53acf3 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=156 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=156 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..d54516d54f29052dbd797cbdf12eca4e0806deb8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_16/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:08,432][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:16,903][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:17,059][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:17,059][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:17,059][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:17,059][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:20,528][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0510564e7cd62a2e332c13ad0d776d2dadf78216 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 165 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 165 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..6fea9ca68df080c3b770dc8f16f69db1f95c2e09 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=165 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=165 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=165,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=165,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7f23710702264edb6d223bac0538861d6c4b94f6 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=165 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=165 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..598c904b283fa438700ff4984290546277f5717d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_17/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:04,444][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:12,901][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:13,041][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:13,041][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:13,041][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:13,041][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:16,300][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..600f07cd2557722a175aea0a6046d1fb75099734 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 174 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 174 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..6e9eb1f9637e64c8d18e60503f2a5f5e5e35147e --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=174 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=174 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=174,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=174,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..2982e40065f578bba7e49bf1eb6e5f3031e6b775 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=174 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=174 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..8948eeb6c60ffc0709ef79919f39f20f58c1bed6 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_18/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:01,283][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:09,595][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:09,739][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:09,739][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:09,739][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:09,739][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:14,760][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4e0a56968db2ba2004d516aba928cc1d3a209c39 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 183 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 183 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d9c520752189ebea7495d32906a73ab271e8a208 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=183 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=183 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=183,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=183,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..39a904561b5884d2bf874f7bf8c4f77ce7e22223 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=183 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=183 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..e3d3b14c882f2f38dc67ee1c46e33757b1c44198 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_19/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:06,548][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:14,892][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:15,035][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:15,035][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:15,035][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:15,035][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:18,565][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9942cb41e38fd29e6ded3761c48775246ff3e6b2 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 192 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 192 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3cfd239d475532376bd2dfcd9fa222a9a8f51727 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=192 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=192 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=192,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=192,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5cdac596b4807a0686699c8a2a2fbafc0d99ee64 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=192 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=192 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..4e0c04d78a1f5b885e820e1b746ef0fa034f9e08 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_20/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:03:59,095][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:07,428][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:07,564][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:07,564][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:07,565][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:07,565][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:12,277][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..38ced664a3ccd8f40557b78318d639f26763d63d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 201 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 201 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..106fc261707b317d95155bee04724d8269d7da3e --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=201 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=201 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=201,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=201,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d928f700d5501cc298252f9f1c30b4901c338b05 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=201 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=201 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..3301ecc5d15f28fcf9c881f77076dd69581d1912 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_21/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:19,850][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:28,200][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:28,352][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:28,352][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:28,352][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:28,352][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:31,319][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0b5798e312395f190c9b3c566ee7f2465645d05d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 210 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 210 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..979dea05b8c57de1017943c18507239bd2ca8e48 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=210 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=210 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=210,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=210,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5280fc3160ce23261982e2eab1d18000114bb917 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=210 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=210 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..1f5dcb8dce5e0710871424c1859b9b65eadf7205 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_22/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:14,670][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:23,090][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:23,221][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:23,221][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:23,221][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:23,221][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:26,371][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5d6693586d5d7a9731db169f8b009fd3bc8d4b62 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 219 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 219 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f99b4dbb788390808e32454026c42e218c3a8e45 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=219 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=219 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=219,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=219,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d73b2eebcad472eebbd2803abd9fcbac6ebf4ecd --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=219 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=219 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..2f8e471e6684a8c4702e025fcbcda6677422c7e1 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_23/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:06,821][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:15,229][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:15,366][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:15,366][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:15,366][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:15,366][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:19,304][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3d7d407643ee2af253f2d51feabdd3d7caf75164 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 228 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 228 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7c802550182e00bea63940342c00df7d8e12a3ac --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=228 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=228 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=228,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=228,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..2f24ee45bda2056f6c9559a16bfa2a1f3278a862 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=228 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=228 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..7732aaaa73dfe5d4ae12d6d97604e4e647cb439e --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_24/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:04,717][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:13,032][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:13,168][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:13,169][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:13,169][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:13,169][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:16,529][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0c400f9be5f35d043eb6e7b9c7653c74ac17db0d --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 237 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 237 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..557915a65a1276b773be2d4eab847af16868b1b5 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=237 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=237 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=237,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=237,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4b2cde14c9437e3d17d59f49725482107832b77f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=237 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=237 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..28cac84a580c06daa22dbd01b261ed1d041b0439 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_25/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:25,316][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:33,771][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:33,907][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:33,908][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:33,908][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:33,908][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:37,128][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..dc0569bf2ecb1ebee6e7427643d1ee83b5c929f9 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 246 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 246 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b4bca74b81bdcda947886617c5f4af2df50ab05a --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=246 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=246 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=246,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=246,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a607155b3a19a7eb1fb3fcbfacd138086e7cae17 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=246 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=246 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..9c4522567d6585905003d216e98291b6d14d0740 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_26/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:01,050][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:09,350][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:09,496][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:09,497][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:09,497][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:09,497][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:14,177][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0420ebba36d8d9bcd64faed2bce56783ccda87a7 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 255 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 255 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..355422ec789be4fa9017830a73ce8db5ca7320af --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=255 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=255 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=255,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=255,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e61baa5cac9144371a0484b5c282375fbba9c895 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=255 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=255 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..0620b4fc098021c57768eea55a50b995f38b3297 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_27/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:03:59,243][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:07,561][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:07,701][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:07,701][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:07,701][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:07,701][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:13,538][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e85a9e1e2b9228a4ba93fd78952529d6f34e86c1 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 264 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 264 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ec701770dbe3d02f1f16dbb95720a8af91db7c7c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=264 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=264 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=264,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=264,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..af6fda639e46bd57afb8d08b67aa9faae04a62e2 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=264 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=264 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..653658e234379969cb35702129b3fa69906acaff --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_28/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:12,303][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:20,781][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:20,914][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:20,914][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:20,914][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:20,914][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:24,458][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7de868bdd7a8779f21879fa5c90c4cc50a5e986c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 273 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 273 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3e443e66621c77d9c6cc4821bffdc1c51bf60122 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=273 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=273 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=273,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=273,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..6ec12dff5b4c29a220e9bb21d37c5a8e5f1d5339 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=273 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=273 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..95a752ae1940649d8f3806aee3abe8d7d3365e25 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_29/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:03:59,371][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:07,881][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:08,049][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:08,049][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:08,049][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:08,049][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:14,110][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ddfacece93fd5f4648500bb3c5607503fb293b72 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 282 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 282 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..401dde9bbea91239d98a1b2f72847210804dd91a --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=282 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=282 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=282,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=282,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b817aaaec49fddee097692472149849531c52984 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=282 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=282 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..cac1c7f39616cd4cb62147c87e4114de1d547ff9 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_30/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:23,981][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:32,345][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:32,489][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:32,490][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:32,490][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:32,490][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:35,654][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31/hydra/run/.hydra/config.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31/hydra/run/.hydra/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..2b89dfcdf6631a1b33ccf08880151436a2ac2bf7 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31/hydra/run/.hydra/config.yaml @@ -0,0 +1,247 @@ +experiment: + debug: ${debug} + tasks: + - test + num_nodes: 1 + training: + precision: 16-mixed + compile: false + lr: 2.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 2000005 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + save_last_k: null + enable_version_counter: false + curriculum: + enabled: true + stages: + - name: no_updown + until_step: 240000 + dataset: + wo_updown: true + - name: with_updown + until_step: ${experiment.training.max_steps} + dataset: + wo_updown: false + validation: + precision: 16-mixed + compile: false + batch_size: 4 + val_every_n_step: 2500 + val_every_n_epoch: null + limit_batch: 1 + inference_mode: false + data: + num_workers: 4 + shuffle: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 9 + inference_mode: false + data: + num_workers: 2 + shuffle: false + manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + manifest_start: 291 + manifest_count: 9 + logging: + metrics: null +dataset: + debug: ${debug} + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + metadata: null + n_frames: 8 + n_frames_valid: 1100 + frame_skip: 1 + resolution: 128 + image_hw: + - 360 + - 640 + vertical_fov_degrees: 70.0 + observation_shape: + - 16 + - 18 + - 32 + data_mean: 0.0 + data_std: 1.0 + action_cond_dim: 25 + context_length: 100 + single_eval_clip: true + shuffle_clips: true + memory_selection: + enabled: true + image_hw: ${..image_hw} + vertical_fov_degrees: ${..vertical_fov_degrees} + causal: true + max_anchor_frames: 2 + max_dynamic_frames: 4 + max_revisit_frames: 2 + training_dropout: 0.1 + fov_overlap_threshold: 0.6 + min_total_selected_coverage: 0.1 + local_context_exclusion_frames: 8 + plucker_moment_radius: 30.0 + anchor_diverse_selection: true + pose_preselect_topk: 32 + candidate_chunk_size: 0 + dynamic: + selection_policy: multiview + multiview_selector: shared_area_diverse + shared_area_threshold: 0.3 + min_pose_novelty: 0.1 + shortlist_per_reference: 16 + scene_threshold: 2.5 + state_threshold: 2.5 + stable_threshold: 1.0 + stable_frames: 3 + min_event_gap: 8 + min_anchor_score: 2.0 + max_event_anchors: null + b_pose: 0.3 + b_action: 0.2 + _name: video_minecraft_dememwm_latent + seed: 42 + wo_updown: false +algorithm: + debug: ${debug} + lr: ${experiment.training.lr} + x_shape: ${dataset.observation_shape} + frame_stack: 1 + frame_skip: ${dataset.frame_skip} + data_mean: ${dataset.data_mean} + data_std: ${dataset.data_std} + external_cond_dim: 0 + context_frames: 600 + weight_decay: 0.002 + warmup_steps: 1000 + optimizer_beta: + - 0.9 + - 0.99 + uncertainty_scale: 1 + guidance_scale: 0.0 + chunk_size: 1 + scheduling_matrix: autoregressive + noise_level: random_all + causal: true + diffusion: + objective: pred_v + beta_schedule: sigmoid + schedule_fn_kwargs: {} + clip_noise: 20.0 + use_snr: false + use_cum_snr: false + use_fused_snr: true + snr_clip: 5.0 + cum_snr_decay: 0.96 + timesteps: 1000 + sampling_timesteps: 20 + ddim_sampling_eta: 0.0 + stabilization_level: 15 + architecture: + network_size: 64 + attn_heads: 4 + attn_dim_head: 64 + dim_mults: + - 1 + - 2 + - 4 + - 8 + resolution: ${dataset.resolution} + attn_resolutions: + - 16 + - 32 + - 64 + - 128 + use_init_temporal_attn: true + use_linear_attn: true + time_emb_type: rotary + n_frames: ${dataset.n_frames} + metadata: ${dataset.metadata} + memory_selection: ${dataset.memory_selection} + memory_source_simulation: + training_generated_suffix_max_frames: null + memory_noise: + enabled: true + mode: source_free_logsnr + stable_probability: 0.25 + noisy_distribution: query_to_stabilization_uniform_logsnr_v1 + dynamic_generated_max_fraction: 0.5 + revisit_generated_max_fraction: 0.5 + anchor_max_fraction: 0.25 + dynamic_max_fraction: 0.5 + revisit_max_fraction: 0.25 + validation_noisy_memory: false + noise_route: + anchor: high + dynamic: low + revisit: all + trainability: + freeze_vae: true + reference_attention: true + adaln_mlp: true + geometry_projections: true + train_full_dit: false + full_dit_start_step: null + lr: + memory_modules: 8.0e-05 + base_dit: 2.0e-05 + n_tokens: 8 + action_cond_dim: ${dataset.action_cond_dim} + pose_cond_dim: 5 + use_plucker: true + relative_embedding: true + state_embed_only_on_qk: true + use_memory_attention: true + unified_memory_attention: true + anchor_zero_geometry: false + memory_role_embedding: true + memory_source_embedding: false + memory_attention_key_only_geometry: true + memory_age: + enabled: true + num_buckets: 32 + num_exact_buckets: 17 + overflow_start: 100 + memory_stream_count_correction: false + ref_mode: sequential + vertical_fov_degrees: ${dataset.vertical_fov_degrees} + log_video: false + save_local: true + metric_report_segment: 100 + require_pose_prediction: false + _name: dememwm_base + local_save_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31 + lpips_batch_size: 16 + eval_manifest_path: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + eval_manifest_start: 291 + eval_manifest_count: 9 + eval_master_seed: 43 + eval_batch_size: 1 +debug: false +wandb: + entity: turlin + project: worldmem + mode: disabled +resume: null +load: /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +auto_resume: false +resume_ckpt_path: null +allow_dememwm_legacy_migration: false +load_ema_weights: false +name: dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +customized_load: false +seed: 43 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31/hydra/run/.hydra/hydra.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31/hydra/run/.hydra/hydra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c184fee2b18a5241fca7f95d6cc7ec539b708038 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31/hydra/run/.hydra/hydra.yaml @@ -0,0 +1,227 @@ +hydra: + run: + dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31/hydra/run + sweep: + dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S} + subdir: ${hydra.job.num} + launcher: + _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher + sweeper: + _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper + max_batch_size: null + params: null + help: + app_name: ${hydra.job.name} + header: '${hydra.help.app_name} is powered by Hydra. + + ' + footer: 'Powered by Hydra (https://hydra.cc) + + Use --hydra-help to view Hydra specific help + + ' + template: '${hydra.help.header} + + == Configuration groups == + + Compose your configuration from those groups (group=option) + + + $APP_CONFIG_GROUPS + + + == Config == + + Override anything in the config (foo.bar=value) + + + $CONFIG + + + ${hydra.help.footer} + + ' + hydra_help: + template: 'Hydra (${hydra.runtime.version}) + + See https://hydra.cc for more info. + + + == Flags == + + $FLAGS_HELP + + + == Configuration groups == + + Compose your configuration from those groups (For example, append hydra/job_logging=disabled + to command line) + + + $HYDRA_CONFIG_GROUPS + + + Use ''--cfg hydra'' to Show the Hydra config. + + ' + hydra_help: ??? + hydra_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][HYDRA] %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + root: + level: INFO + handlers: + - console + loggers: + logging_example: + level: DEBUG + disable_existing_loggers: false + job_logging: + version: 1 + formatters: + simple: + format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s' + handlers: + console: + class: logging.StreamHandler + formatter: simple + stream: ext://sys.stdout + file: + class: logging.FileHandler + formatter: simple + filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log + root: + level: INFO + handlers: + - console + - file + disable_existing_loggers: false + env: {} + mode: RUN + searchpath: [] + callbacks: {} + output_subdir: .hydra + overrides: + hydra: + - hydra.run.dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31/hydra/run + - hydra.mode=RUN + task: + - +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 + - load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt + - auto_resume=false + - +customized_load=false + - wandb.mode=disabled + - experiment.tasks=[test] + - experiment.num_nodes=1 + - +seed=43 + - dataset=video_minecraft_dememwm_latent + - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft + - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features + - dataset.n_frames=8 + - dataset.n_frames_valid=1100 + - dataset.context_length=100 + - dataset.single_eval_clip=true + - dataset.memory_selection.enabled=true + - dataset.memory_selection.causal=true + - dataset.memory_selection.max_anchor_frames=2 + - dataset.memory_selection.max_dynamic_frames=4 + - dataset.memory_selection.max_revisit_frames=2 + - dataset.memory_selection.training_dropout=0.1 + - dataset.memory_selection.dynamic.selection_policy=multiview + - dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse + - +dataset.seed=42 + - +dataset.wo_updown=false + - algorithm=dememwm_base + - algorithm.causal=true + - algorithm.n_tokens=8 + - algorithm.context_frames=600 + - algorithm.metric_report_segment=100 + - algorithm.use_memory_attention=true + - algorithm.anchor_zero_geometry=false + - algorithm.memory_role_embedding=true + - algorithm.memory_source_embedding=false + - algorithm.memory_attention_key_only_geometry=true + - algorithm.memory_age.enabled=true + - algorithm.memory_age.num_buckets=32 + - algorithm.memory_age.num_exact_buckets=17 + - algorithm.memory_age.overflow_start=100 + - algorithm.memory_stream_count_correction=false + - algorithm.noise_route.anchor=high + - algorithm.noise_route.dynamic=low + - algorithm.noise_route.revisit=all + - algorithm.memory_source_simulation.training_generated_suffix_max_frames=null + - algorithm.memory_noise.enabled=true + - algorithm.memory_noise.mode=source_free_logsnr + - algorithm.memory_noise.stable_probability=0.25 + - algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 + - algorithm.memory_noise.validation_noisy_memory=false + - algorithm.chunk_size=1 + - algorithm.diffusion.sampling_timesteps=20 + - algorithm.diffusion.ddim_sampling_eta=0.0 + - algorithm.log_video=false + - algorithm.save_local=true + - +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31 + - +algorithm.lpips_batch_size=16 + - +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +algorithm.eval_manifest_start=291 + - +algorithm.eval_manifest_count=9 + - +algorithm.eval_master_seed=43 + - +algorithm.eval_batch_size=1 + - +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv + - +experiment.test.manifest_start=291 + - +experiment.test.manifest_count=9 + - experiment.test.batch_size=1 + - experiment.test.limit_batch=9 + - experiment.test.data.num_workers=2 + - experiment.test.inference_mode=false + job: + name: main + chdir: null + override_dirname: +algorithm.eval_batch_size=1,+algorithm.eval_manifest_count=9,+algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+algorithm.eval_manifest_start=291,+algorithm.eval_master_seed=43,+algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31,+algorithm.lpips_batch_size=16,+customized_load=false,+dataset.seed=42,+dataset.wo_updown=false,+experiment.test.manifest_count=9,+experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv,+experiment.test.manifest_start=291,+name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43,+seed=43,algorithm.anchor_zero_geometry=false,algorithm.causal=true,algorithm.chunk_size=1,algorithm.context_frames=600,algorithm.diffusion.ddim_sampling_eta=0.0,algorithm.diffusion.sampling_timesteps=20,algorithm.log_video=false,algorithm.memory_age.enabled=true,algorithm.memory_age.num_buckets=32,algorithm.memory_age.num_exact_buckets=17,algorithm.memory_age.overflow_start=100,algorithm.memory_attention_key_only_geometry=true,algorithm.memory_noise.enabled=true,algorithm.memory_noise.mode=source_free_logsnr,algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1,algorithm.memory_noise.stable_probability=0.25,algorithm.memory_noise.validation_noisy_memory=false,algorithm.memory_role_embedding=true,algorithm.memory_source_embedding=false,algorithm.memory_source_simulation.training_generated_suffix_max_frames=null,algorithm.memory_stream_count_correction=false,algorithm.metric_report_segment=100,algorithm.n_tokens=8,algorithm.noise_route.anchor=high,algorithm.noise_route.dynamic=low,algorithm.noise_route.revisit=all,algorithm.save_local=true,algorithm.use_memory_attention=true,algorithm=dememwm_base,auto_resume=false,dataset.context_length=100,dataset.memory_selection.causal=true,dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse,dataset.memory_selection.dynamic.selection_policy=multiview,dataset.memory_selection.enabled=true,dataset.memory_selection.max_anchor_frames=2,dataset.memory_selection.max_dynamic_frames=4,dataset.memory_selection.max_revisit_frames=2,dataset.memory_selection.training_dropout=0.1,dataset.n_frames=8,dataset.n_frames_valid=1100,dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features,dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft,dataset.single_eval_clip=true,dataset=video_minecraft_dememwm_latent,experiment.num_nodes=1,experiment.tasks=[test],experiment.test.batch_size=1,experiment.test.data.num_workers=2,experiment.test.inference_mode=false,experiment.test.limit_batch=9,load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt,wandb.mode=disabled + id: ??? + num: ??? + config_name: training + env_set: {} + env_copy: [] + config: + override_dirname: + kv_sep: '=' + item_sep: ',' + exclude_keys: [] + runtime: + version: 1.3.2 + version_base: '1.3' + cwd: /proj/cvl/users/x_fahkh2/DeMemWM + config_sources: + - path: hydra.conf + schema: pkg + provider: hydra + - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations + schema: file + provider: main + - path: '' + schema: structured + provider: schema + output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31/hydra/run + choices: + cluster: null + algorithm: dememwm_base + dataset: video_minecraft_dememwm_latent + experiment: exp_video + hydra/env: default + hydra/callbacks: null + hydra/job_logging: default + hydra/hydra_logging: default + hydra/hydra_help: default + hydra/help: default + hydra/sweeper: basic + hydra/launcher: basic + hydra/output: default + verbose: false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31/hydra/run/.hydra/overrides.yaml b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31/hydra/run/.hydra/overrides.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a107761ba3da8f4f3be6a7be294967352729b67e --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31/hydra/run/.hydra/overrides.yaml @@ -0,0 +1,68 @@ +- +name=dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43 +- load=/proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt +- auto_resume=false +- +customized_load=false +- wandb.mode=disabled +- experiment.tasks=[test] +- experiment.num_nodes=1 +- +seed=43 +- dataset=video_minecraft_dememwm_latent +- dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft +- dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features +- dataset.n_frames=8 +- dataset.n_frames_valid=1100 +- dataset.context_length=100 +- dataset.single_eval_clip=true +- dataset.memory_selection.enabled=true +- dataset.memory_selection.causal=true +- dataset.memory_selection.max_anchor_frames=2 +- dataset.memory_selection.max_dynamic_frames=4 +- dataset.memory_selection.max_revisit_frames=2 +- dataset.memory_selection.training_dropout=0.1 +- dataset.memory_selection.dynamic.selection_policy=multiview +- dataset.memory_selection.dynamic.multiview_selector=shared_area_diverse +- +dataset.seed=42 +- +dataset.wo_updown=false +- algorithm=dememwm_base +- algorithm.causal=true +- algorithm.n_tokens=8 +- algorithm.context_frames=600 +- algorithm.metric_report_segment=100 +- algorithm.use_memory_attention=true +- algorithm.anchor_zero_geometry=false +- algorithm.memory_role_embedding=true +- algorithm.memory_source_embedding=false +- algorithm.memory_attention_key_only_geometry=true +- algorithm.memory_age.enabled=true +- algorithm.memory_age.num_buckets=32 +- algorithm.memory_age.num_exact_buckets=17 +- algorithm.memory_age.overflow_start=100 +- algorithm.memory_stream_count_correction=false +- algorithm.noise_route.anchor=high +- algorithm.noise_route.dynamic=low +- algorithm.noise_route.revisit=all +- algorithm.memory_source_simulation.training_generated_suffix_max_frames=null +- algorithm.memory_noise.enabled=true +- algorithm.memory_noise.mode=source_free_logsnr +- algorithm.memory_noise.stable_probability=0.25 +- algorithm.memory_noise.noisy_distribution=query_to_stabilization_uniform_logsnr_v1 +- algorithm.memory_noise.validation_noisy_memory=false +- algorithm.chunk_size=1 +- algorithm.diffusion.sampling_timesteps=20 +- algorithm.diffusion.ddim_sampling_eta=0.0 +- algorithm.log_video=false +- algorithm.save_local=true +- +algorithm.local_save_dir=/proj/cvl/users/x_fahkh2/DeMemWM/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31 +- +algorithm.lpips_batch_size=16 +- +algorithm.eval_manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +algorithm.eval_manifest_start=291 +- +algorithm.eval_manifest_count=9 +- +algorithm.eval_master_seed=43 +- +algorithm.eval_batch_size=1 +- +experiment.test.manifest_path=/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/minecraft_test_300.csv +- +experiment.test.manifest_start=291 +- +experiment.test.manifest_count=9 +- experiment.test.batch_size=1 +- experiment.test.limit_batch=9 +- experiment.test.data.num_workers=2 +- experiment.test.inference_mode=false diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31/hydra/run/main.log b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31/hydra/run/main.log new file mode 100644 index 0000000000000000000000000000000000000000..67fa78123054b98e48a3bbaddb30e6b99a2d0264 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_countcorr_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/workers/worker_31/hydra/run/main.log @@ -0,0 +1,7 @@ +[2026-07-22 14:04:13,226][lightning.fabric.utilities.seed][INFO] - [rank: 0] Seed set to 43 +[2026-07-22 14:04:21,648][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP) +[2026-07-22 14:04:21,798][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True +[2026-07-22 14:04:21,798][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores +[2026-07-22 14:04:21,798][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs +[2026-07-22 14:04:21,798][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs +[2026-07-22 14:04:25,339][pytorch_lightning.utilities.rank_zero][INFO] - Restoring states from the checkpoint path at /proj/cvl/users/x_fahkh2/DeMemWM/results/train_dememwm_sourcefree_countcorr_off_8h200_bs16_global128_seed42_200k/checkpoints/epoch3_step200000.ckpt diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/evaluation.patch b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/evaluation.patch new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..3cfd08bc49f4159d2bd5cace6409ac92956d78e1 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.csv @@ -0,0 +1,51 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.sha256 b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..67b0dc2042229c0768d324282215bdf64bebe28f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.sha256 @@ -0,0 +1 @@ +f9536911e434610cce23306ff99575db53cd7fc29e3df841da4b6631bf418634 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/git_provenance.json b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/git_provenance.json new file mode 100644 index 0000000000000000000000000000000000000000..fabe691fc08806f957f2445d4a52b2f38d7f07d8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/git_provenance.json @@ -0,0 +1,5 @@ +{ + "branch": "dememwm-unified-sourcefree-age-balanced", + "head": "732555281e86c4fdbb71c98a000d648c3edc3109", + "status": "?? results/\n?? train_dememwm_full_berzelius_gate_off_h200.sh\n?? train_dememwm_separate_mem_gatefix_4layers512_valid500_berzelius_h200.sh" +} diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..dae70b4071c2d2a834268c2a26f061d308f11f2c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.csv @@ -0,0 +1,301 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +50,50,212,test/val_savanna/1/000039.mp4,test/val_savanna/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +51,51,4,test/val_desert/0/000001.mp4,test/val_desert/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +52,52,7,test/val_desert/0/000002.mp4,test/val_desert/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +53,53,310,test/val_iceplains/0/000026.mp4,test/val_iceplains/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +54,54,284,test/val_iceplains/0/000015.mp4,test/val_iceplains/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +55,55,109,test/val_desert/3/000041.mp4,test/val_desert/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +56,56,45,test/val_desert/0/000018.mp4,test/val_desert/0/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +57,57,93,test/val_desert/0/000035.mp4,test/val_desert/0/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +58,58,400,test/val_plains/3/000011.mp4,test/val_plains/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +59,59,300,test/val_iceplains/2/000021.mp4,test/val_iceplains/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +60,60,226,test/val_savanna/2/000043.mp4,test/val_savanna/2/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +61,61,417,test/val_plains/2/000018.mp4,test/val_plains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +62,62,120,test/val_desert/3/000046.mp4,test/val_desert/3/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +63,63,106,test/val_desert/3/000040.mp4,test/val_desert/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +64,64,345,test/val_iceplains/1/000039.mp4,test/val_iceplains/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +65,65,169,test/val_savanna/2/000016.mp4,test/val_savanna/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +66,66,312,test/val_iceplains/1/000027.mp4,test/val_iceplains/1/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +67,67,368,test/val_iceplains/0/000048.mp4,test/val_iceplains/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +68,68,449,test/val_plains/1/000034.mp4,test/val_plains/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +69,69,11,test/val_desert/1/000004.mp4,test/val_desert/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +70,70,6,test/val_desert/2/000001.mp4,test/val_desert/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +71,71,257,test/val_iceplains/0/000005.mp4,test/val_iceplains/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +72,72,262,test/val_iceplains/3/000006.mp4,test/val_iceplains/3/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +73,73,164,test/val_savanna/0/000015.mp4,test/val_savanna/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +74,74,337,test/val_iceplains/2/000036.mp4,test/val_iceplains/2/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +75,75,95,test/val_desert/0/000036.mp4,test/val_desert/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +76,76,277,test/val_iceplains/0/000013.mp4,test/val_iceplains/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +77,77,26,test/val_desert/1/000010.mp4,test/val_desert/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +78,78,77,test/val_desert/0/000029.mp4,test/val_desert/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +79,79,294,test/val_iceplains/1/000019.mp4,test/val_iceplains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +80,80,34,test/val_desert/1/000014.mp4,test/val_desert/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +81,81,188,test/val_savanna/1/000029.mp4,test/val_savanna/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +82,82,170,test/val_savanna/0/000017.mp4,test/val_savanna/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +83,83,415,test/val_plains/2/000017.mp4,test/val_plains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +84,84,145,test/val_savanna/0/000006.mp4,test/val_savanna/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +85,85,133,test/val_savanna/1/000002.mp4,test/val_savanna/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +86,86,347,test/val_iceplains/2/000039.mp4,test/val_iceplains/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +87,87,209,test/val_savanna/0/000038.mp4,test/val_savanna/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +88,88,404,test/val_plains/1/000013.mp4,test/val_plains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +89,89,211,test/val_savanna/2/000038.mp4,test/val_savanna/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +90,90,464,test/val_plains/2/000044.mp4,test/val_plains/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +91,91,162,test/val_savanna/1/000014.mp4,test/val_savanna/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +92,92,419,test/val_plains/2/000019.mp4,test/val_plains/2/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +93,93,127,test/val_desert/3/000049.mp4,test/val_desert/3/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +94,94,370,test/val_iceplains/2/000048.mp4,test/val_iceplains/2/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +95,95,293,test/val_iceplains/2/000018.mp4,test/val_iceplains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +96,96,251,test/val_iceplains/1/000003.mp4,test/val_iceplains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +97,97,441,test/val_plains/1/000031.mp4,test/val_plains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +98,98,69,test/val_desert/2/000026.mp4,test/val_desert/2/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +99,99,286,test/val_iceplains/2/000015.mp4,test/val_iceplains/2/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +100,100,438,test/val_plains/0/000029.mp4,test/val_plains/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +101,101,376,test/val_plains/0/000002.mp4,test/val_plains/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +102,102,163,test/val_savanna/2/000014.mp4,test/val_savanna/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +103,103,31,test/val_desert/1/000013.mp4,test/val_desert/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +104,104,241,test/val_savanna/0/000049.mp4,test/val_savanna/0/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +105,105,458,test/val_plains/0/000040.mp4,test/val_plains/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +106,106,371,test/val_iceplains/1/000049.mp4,test/val_iceplains/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +107,107,230,test/val_savanna/3/000045.mp4,test/val_savanna/3/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +108,108,372,test/val_iceplains/2/000049.mp4,test/val_iceplains/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +109,109,223,test/val_savanna/1/000042.mp4,test/val_savanna/1/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +110,110,68,test/val_desert/0/000026.mp4,test/val_desert/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +111,111,461,test/val_plains/2/000042.mp4,test/val_plains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +112,112,151,test/val_savanna/1/000009.mp4,test/val_savanna/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +113,113,72,test/val_desert/3/000027.mp4,test/val_desert/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +114,114,427,test/val_plains/0/000024.mp4,test/val_plains/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +115,115,317,test/val_iceplains/3/000029.mp4,test/val_iceplains/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +116,116,249,test/val_iceplains/2/000001.mp4,test/val_iceplains/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +117,117,364,test/val_iceplains/2/000046.mp4,test/val_iceplains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +118,118,96,test/val_desert/2/000037.mp4,test/val_desert/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +119,119,190,test/val_savanna/3/000029.mp4,test/val_savanna/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +120,120,227,test/val_savanna/2/000044.mp4,test/val_savanna/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +121,121,244,test/val_iceplains/0/000000.mp4,test/val_iceplains/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +122,122,38,test/val_desert/3/000015.mp4,test/val_desert/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +123,123,291,test/val_iceplains/2/000017.mp4,test/val_iceplains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +124,124,85,test/val_desert/2/000032.mp4,test/val_desert/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +125,125,358,test/val_iceplains/0/000044.mp4,test/val_iceplains/0/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +126,126,50,test/val_desert/1/000020.mp4,test/val_desert/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +127,127,207,test/val_savanna/0/000036.mp4,test/val_savanna/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +128,128,330,test/val_iceplains/2/000034.mp4,test/val_iceplains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +129,129,132,test/val_savanna/0/000001.mp4,test/val_savanna/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +130,130,423,test/val_plains/3/000021.mp4,test/val_plains/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +131,131,66,test/val_desert/2/000025.mp4,test/val_desert/2/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +132,132,153,test/val_savanna/3/000009.mp4,test/val_savanna/3/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +133,133,382,test/val_plains/0/000004.mp4,test/val_plains/0/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +134,134,143,test/val_savanna/2/000005.mp4,test/val_savanna/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +135,135,348,test/val_iceplains/1/000040.mp4,test/val_iceplains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +136,136,384,test/val_plains/2/000005.mp4,test/val_plains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +137,137,30,test/val_desert/0/000012.mp4,test/val_desert/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +138,138,203,test/val_savanna/2/000034.mp4,test/val_savanna/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +139,139,179,test/val_savanna/0/000022.mp4,test/val_savanna/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +140,140,253,test/val_iceplains/2/000003.mp4,test/val_iceplains/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +141,141,289,test/val_iceplains/2/000016.mp4,test/val_iceplains/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +142,142,422,test/val_plains/2/000020.mp4,test/val_plains/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +143,143,20,test/val_desert/0/000008.mp4,test/val_desert/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +144,144,130,test/val_savanna/0/000000.mp4,test/val_savanna/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +145,145,296,test/val_iceplains/1/000020.mp4,test/val_iceplains/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +146,146,278,test/val_iceplains/3/000013.mp4,test/val_iceplains/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +147,147,269,test/val_iceplains/0/000009.mp4,test/val_iceplains/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +148,148,314,test/val_iceplains/0/000028.mp4,test/val_iceplains/0/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +149,149,281,test/val_iceplains/0/000014.mp4,test/val_iceplains/0/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +150,150,149,test/val_savanna/0/000008.mp4,test/val_savanna/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +151,151,184,test/val_savanna/0/000024.mp4,test/val_savanna/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +152,152,407,test/val_plains/1/000014.mp4,test/val_plains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +153,153,304,test/val_iceplains/2/000023.mp4,test/val_iceplains/2/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +154,154,208,test/val_savanna/3/000037.mp4,test/val_savanna/3/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +155,155,182,test/val_savanna/3/000023.mp4,test/val_savanna/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +156,156,394,test/val_plains/1/000009.mp4,test/val_plains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +157,157,14,test/val_desert/0/000006.mp4,test/val_desert/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +158,158,55,test/val_desert/3/000021.mp4,test/val_desert/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +159,159,75,test/val_desert/2/000028.mp4,test/val_desert/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +160,160,319,test/val_iceplains/1/000031.mp4,test/val_iceplains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +161,161,39,test/val_desert/1/000016.mp4,test/val_desert/1/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +162,162,418,test/val_plains/1/000019.mp4,test/val_plains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +163,163,389,test/val_plains/3/000007.mp4,test/val_plains/3/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +164,164,409,test/val_plains/2/000014.mp4,test/val_plains/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +165,165,344,test/val_iceplains/2/000038.mp4,test/val_iceplains/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +166,166,396,test/val_plains/2/000009.mp4,test/val_plains/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +167,167,67,test/val_desert/1/000026.mp4,test/val_desert/1/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +168,168,99,test/val_desert/3/000038.mp4,test/val_desert/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +169,169,288,test/val_iceplains/3/000016.mp4,test/val_iceplains/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +170,170,97,test/val_desert/1/000038.mp4,test/val_desert/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +171,171,178,test/val_savanna/2/000021.mp4,test/val_savanna/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +172,172,264,test/val_iceplains/1/000007.mp4,test/val_iceplains/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +173,173,266,test/val_iceplains/0/000008.mp4,test/val_iceplains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +174,174,303,test/val_iceplains/1/000023.mp4,test/val_iceplains/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +175,175,442,test/val_plains/0/000031.mp4,test/val_plains/0/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +176,176,233,test/val_savanna/1/000047.mp4,test/val_savanna/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +177,177,240,test/val_savanna/1/000049.mp4,test/val_savanna/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +178,178,42,test/val_desert/0/000017.mp4,test/val_desert/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +179,179,369,test/val_iceplains/3/000048.mp4,test/val_iceplains/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +180,180,398,test/val_plains/3/000010.mp4,test/val_plains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +181,181,435,test/val_plains/2/000027.mp4,test/val_plains/2/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +182,182,242,test/val_savanna/2/000049.mp4,test/val_savanna/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +183,183,361,test/val_iceplains/2/000045.mp4,test/val_iceplains/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +184,184,470,test/val_plains/2/000047.mp4,test/val_plains/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +185,185,356,test/val_iceplains/2/000042.mp4,test/val_iceplains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +186,186,468,test/val_plains/2/000046.mp4,test/val_plains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +187,187,113,test/val_desert/0/000043.mp4,test/val_desert/0/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +188,188,19,test/val_desert/1/000008.mp4,test/val_desert/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +189,189,199,test/val_savanna/0/000033.mp4,test/val_savanna/0/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +190,190,118,test/val_desert/2/000045.mp4,test/val_desert/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +191,191,433,test/val_plains/0/000027.mp4,test/val_plains/0/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +192,192,328,test/val_iceplains/0/000034.mp4,test/val_iceplains/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +193,193,213,test/val_savanna/0/000039.mp4,test/val_savanna/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +194,194,339,test/val_iceplains/0/000037.mp4,test/val_iceplains/0/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +195,195,466,test/val_plains/1/000046.mp4,test/val_plains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +196,196,355,test/val_iceplains/3/000042.mp4,test/val_iceplains/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +197,197,336,test/val_iceplains/3/000036.mp4,test/val_iceplains/3/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +198,198,420,test/val_plains/0/000020.mp4,test/val_plains/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +199,199,410,test/val_plains/1/000015.mp4,test/val_plains/1/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +200,200,27,test/val_desert/1/000011.mp4,test/val_desert/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +201,201,86,test/val_desert/1/000033.mp4,test/val_desert/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +202,202,210,test/val_savanna/3/000038.mp4,test/val_savanna/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +203,203,103,test/val_desert/3/000039.mp4,test/val_desert/3/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +204,204,245,test/val_iceplains/3/000000.mp4,test/val_iceplains/3/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +205,205,25,test/val_desert/2/000009.mp4,test/val_desert/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +206,206,340,test/val_iceplains/2/000037.mp4,test/val_iceplains/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +207,207,167,test/val_savanna/0/000016.mp4,test/val_savanna/0/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +208,208,424,test/val_plains/0/000022.mp4,test/val_plains/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +209,209,448,test/val_plains/2/000033.mp4,test/val_plains/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +210,210,386,test/val_plains/0/000006.mp4,test/val_plains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +211,211,392,test/val_plains/3/000008.mp4,test/val_plains/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +212,212,105,test/val_desert/1/000040.mp4,test/val_desert/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +213,213,267,test/val_iceplains/2/000008.mp4,test/val_iceplains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +214,214,90,test/val_desert/0/000034.mp4,test/val_desert/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +215,215,426,test/val_plains/0/000023.mp4,test/val_plains/0/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +216,216,168,test/val_savanna/3/000016.mp4,test/val_savanna/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +217,217,218,test/val_savanna/3/000040.mp4,test/val_savanna/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +218,218,445,test/val_plains/0/000032.mp4,test/val_plains/0/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +219,219,48,test/val_desert/3/000019.mp4,test/val_desert/3/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +220,220,104,test/val_desert/2/000039.mp4,test/val_desert/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +221,221,206,test/val_savanna/2/000035.mp4,test/val_savanna/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +222,222,228,test/val_savanna/1/000045.mp4,test/val_savanna/1/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +223,223,200,test/val_savanna/2/000033.mp4,test/val_savanna/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +224,224,62,test/val_desert/2/000024.mp4,test/val_desert/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +225,225,121,test/val_desert/1/000047.mp4,test/val_desert/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +226,226,147,test/val_savanna/0/000007.mp4,test/val_savanna/0/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +227,227,146,test/val_savanna/2/000006.mp4,test/val_savanna/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +228,228,225,test/val_savanna/1/000043.mp4,test/val_savanna/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +229,229,313,test/val_iceplains/3/000027.mp4,test/val_iceplains/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +230,230,354,test/val_iceplains/0/000042.mp4,test/val_iceplains/0/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +231,231,367,test/val_iceplains/1/000048.mp4,test/val_iceplains/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +232,232,297,test/val_iceplains/3/000020.mp4,test/val_iceplains/3/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +233,233,124,test/val_desert/1/000048.mp4,test/val_desert/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +234,234,171,test/val_savanna/2/000017.mp4,test/val_savanna/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +235,235,342,test/val_iceplains/0/000038.mp4,test/val_iceplains/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +236,236,60,test/val_desert/3/000023.mp4,test/val_desert/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +237,237,131,test/val_savanna/1/000001.mp4,test/val_savanna/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +238,238,365,test/val_iceplains/1/000047.mp4,test/val_iceplains/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +239,239,315,test/val_iceplains/2/000028.mp4,test/val_iceplains/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +240,240,220,test/val_savanna/1/000041.mp4,test/val_savanna/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +241,241,8,test/val_desert/1/000003.mp4,test/val_desert/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +242,242,261,test/val_iceplains/0/000006.mp4,test/val_iceplains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +243,243,17,test/val_desert/1/000007.mp4,test/val_desert/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +244,244,462,test/val_plains/1/000043.mp4,test/val_plains/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +245,245,454,test/val_plains/3/000038.mp4,test/val_plains/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +246,246,351,test/val_iceplains/1/000041.mp4,test/val_iceplains/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +247,247,383,test/val_plains/3/000005.mp4,test/val_plains/3/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +248,248,18,test/val_desert/2/000007.mp4,test/val_desert/2/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +249,249,457,test/val_plains/1/000040.mp4,test/val_plains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +250,250,326,test/val_iceplains/3/000033.mp4,test/val_iceplains/3/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +251,251,254,test/val_iceplains/1/000004.mp4,test/val_iceplains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +252,252,375,test/val_plains/1/000002.mp4,test/val_plains/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +253,253,141,test/val_savanna/0/000005.mp4,test/val_savanna/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +254,254,306,test/val_iceplains/2/000024.mp4,test/val_iceplains/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +255,255,10,test/val_desert/2/000003.mp4,test/val_desert/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +256,256,173,test/val_savanna/0/000020.mp4,test/val_savanna/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +257,257,320,test/val_iceplains/1/000032.mp4,test/val_iceplains/1/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +258,258,115,test/val_desert/1/000044.mp4,test/val_desert/1/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +259,259,341,test/val_iceplains/1/000038.mp4,test/val_iceplains/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +260,260,61,test/val_desert/3/000024.mp4,test/val_desert/3/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +261,261,231,test/val_savanna/1/000046.mp4,test/val_savanna/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +262,262,224,test/val_savanna/3/000042.mp4,test/val_savanna/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +263,263,292,test/val_iceplains/3/000018.mp4,test/val_iceplains/3/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +264,264,334,test/val_iceplains/1/000036.mp4,test/val_iceplains/1/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +265,265,159,test/val_savanna/0/000013.mp4,test/val_savanna/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +266,266,102,test/val_desert/0/000039.mp4,test/val_desert/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +267,267,362,test/val_iceplains/1/000046.mp4,test/val_iceplains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +268,268,322,test/val_iceplains/3/000032.mp4,test/val_iceplains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +269,269,381,test/val_plains/1/000004.mp4,test/val_plains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +270,270,84,test/val_desert/3/000032.mp4,test/val_desert/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +271,271,243,test/val_iceplains/1/000000.mp4,test/val_iceplains/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +272,272,65,test/val_desert/3/000025.mp4,test/val_desert/3/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +273,273,64,test/val_desert/0/000025.mp4,test/val_desert/0/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +274,274,272,test/val_iceplains/2/000010.mp4,test/val_iceplains/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +275,275,399,test/val_plains/0/000011.mp4,test/val_plains/0/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +276,276,248,test/val_iceplains/3/000001.mp4,test/val_iceplains/3/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +277,277,43,test/val_desert/3/000017.mp4,test/val_desert/3/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +278,278,393,test/val_plains/2/000008.mp4,test/val_plains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +279,279,123,test/val_desert/2/000047.mp4,test/val_desert/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +280,280,29,test/val_desert/2/000011.mp4,test/val_desert/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +281,281,122,test/val_desert/0/000047.mp4,test/val_desert/0/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +282,282,157,test/val_savanna/3/000011.mp4,test/val_savanna/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +283,283,333,test/val_iceplains/2/000035.mp4,test/val_iceplains/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +284,284,307,test/val_iceplains/1/000025.mp4,test/val_iceplains/1/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +285,285,9,test/val_desert/0/000003.mp4,test/val_desert/0/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +286,286,250,test/val_iceplains/2/000002.mp4,test/val_iceplains/2/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +287,287,165,test/val_savanna/3/000015.mp4,test/val_savanna/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +288,288,0,test/val_desert/1/000000.mp4,test/val_desert/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +289,289,271,test/val_iceplains/3/000010.mp4,test/val_iceplains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +290,290,276,test/val_iceplains/1/000013.mp4,test/val_iceplains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +291,291,82,test/val_desert/1/000031.mp4,test/val_desert/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +292,292,191,test/val_savanna/2/000029.mp4,test/val_savanna/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +293,293,374,test/val_plains/1/000001.mp4,test/val_plains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +294,294,446,test/val_plains/3/000032.mp4,test/val_plains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +295,295,378,test/val_plains/1/000003.mp4,test/val_plains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +296,296,152,test/val_savanna/0/000009.mp4,test/val_savanna/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +297,297,54,test/val_desert/0/000021.mp4,test/val_desert/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +298,298,259,test/val_iceplains/2/000005.mp4,test/val_iceplains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +299,299,91,test/val_desert/2/000034.mp4,test/val_desert/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.sha256 b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..065350069e5b0f46e33cd51ad63f60aa956822fd --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.sha256 @@ -0,0 +1 @@ +ef4e4de9d3975d8a7e724bb609e57f20186cd1f7285457c00fb85c1aca973b99 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/evaluation.patch b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/evaluation.patch new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..3cfd08bc49f4159d2bd5cace6409ac92956d78e1 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.csv @@ -0,0 +1,51 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.sha256 b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..67b0dc2042229c0768d324282215bdf64bebe28f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.sha256 @@ -0,0 +1 @@ +f9536911e434610cce23306ff99575db53cd7fc29e3df841da4b6631bf418634 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/git_provenance.json b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/git_provenance.json new file mode 100644 index 0000000000000000000000000000000000000000..fabe691fc08806f957f2445d4a52b2f38d7f07d8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/git_provenance.json @@ -0,0 +1,5 @@ +{ + "branch": "dememwm-unified-sourcefree-age-balanced", + "head": "732555281e86c4fdbb71c98a000d648c3edc3109", + "status": "?? results/\n?? train_dememwm_full_berzelius_gate_off_h200.sh\n?? train_dememwm_separate_mem_gatefix_4layers512_valid500_berzelius_h200.sh" +} diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..dae70b4071c2d2a834268c2a26f061d308f11f2c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.csv @@ -0,0 +1,301 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +50,50,212,test/val_savanna/1/000039.mp4,test/val_savanna/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +51,51,4,test/val_desert/0/000001.mp4,test/val_desert/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +52,52,7,test/val_desert/0/000002.mp4,test/val_desert/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +53,53,310,test/val_iceplains/0/000026.mp4,test/val_iceplains/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +54,54,284,test/val_iceplains/0/000015.mp4,test/val_iceplains/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +55,55,109,test/val_desert/3/000041.mp4,test/val_desert/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +56,56,45,test/val_desert/0/000018.mp4,test/val_desert/0/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +57,57,93,test/val_desert/0/000035.mp4,test/val_desert/0/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +58,58,400,test/val_plains/3/000011.mp4,test/val_plains/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +59,59,300,test/val_iceplains/2/000021.mp4,test/val_iceplains/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +60,60,226,test/val_savanna/2/000043.mp4,test/val_savanna/2/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +61,61,417,test/val_plains/2/000018.mp4,test/val_plains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +62,62,120,test/val_desert/3/000046.mp4,test/val_desert/3/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +63,63,106,test/val_desert/3/000040.mp4,test/val_desert/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +64,64,345,test/val_iceplains/1/000039.mp4,test/val_iceplains/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +65,65,169,test/val_savanna/2/000016.mp4,test/val_savanna/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +66,66,312,test/val_iceplains/1/000027.mp4,test/val_iceplains/1/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +67,67,368,test/val_iceplains/0/000048.mp4,test/val_iceplains/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +68,68,449,test/val_plains/1/000034.mp4,test/val_plains/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +69,69,11,test/val_desert/1/000004.mp4,test/val_desert/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +70,70,6,test/val_desert/2/000001.mp4,test/val_desert/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +71,71,257,test/val_iceplains/0/000005.mp4,test/val_iceplains/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +72,72,262,test/val_iceplains/3/000006.mp4,test/val_iceplains/3/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +73,73,164,test/val_savanna/0/000015.mp4,test/val_savanna/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +74,74,337,test/val_iceplains/2/000036.mp4,test/val_iceplains/2/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +75,75,95,test/val_desert/0/000036.mp4,test/val_desert/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +76,76,277,test/val_iceplains/0/000013.mp4,test/val_iceplains/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +77,77,26,test/val_desert/1/000010.mp4,test/val_desert/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +78,78,77,test/val_desert/0/000029.mp4,test/val_desert/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +79,79,294,test/val_iceplains/1/000019.mp4,test/val_iceplains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +80,80,34,test/val_desert/1/000014.mp4,test/val_desert/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +81,81,188,test/val_savanna/1/000029.mp4,test/val_savanna/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +82,82,170,test/val_savanna/0/000017.mp4,test/val_savanna/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +83,83,415,test/val_plains/2/000017.mp4,test/val_plains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +84,84,145,test/val_savanna/0/000006.mp4,test/val_savanna/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +85,85,133,test/val_savanna/1/000002.mp4,test/val_savanna/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +86,86,347,test/val_iceplains/2/000039.mp4,test/val_iceplains/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +87,87,209,test/val_savanna/0/000038.mp4,test/val_savanna/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +88,88,404,test/val_plains/1/000013.mp4,test/val_plains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +89,89,211,test/val_savanna/2/000038.mp4,test/val_savanna/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +90,90,464,test/val_plains/2/000044.mp4,test/val_plains/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +91,91,162,test/val_savanna/1/000014.mp4,test/val_savanna/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +92,92,419,test/val_plains/2/000019.mp4,test/val_plains/2/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +93,93,127,test/val_desert/3/000049.mp4,test/val_desert/3/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +94,94,370,test/val_iceplains/2/000048.mp4,test/val_iceplains/2/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +95,95,293,test/val_iceplains/2/000018.mp4,test/val_iceplains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +96,96,251,test/val_iceplains/1/000003.mp4,test/val_iceplains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +97,97,441,test/val_plains/1/000031.mp4,test/val_plains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +98,98,69,test/val_desert/2/000026.mp4,test/val_desert/2/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +99,99,286,test/val_iceplains/2/000015.mp4,test/val_iceplains/2/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +100,100,438,test/val_plains/0/000029.mp4,test/val_plains/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +101,101,376,test/val_plains/0/000002.mp4,test/val_plains/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +102,102,163,test/val_savanna/2/000014.mp4,test/val_savanna/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +103,103,31,test/val_desert/1/000013.mp4,test/val_desert/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +104,104,241,test/val_savanna/0/000049.mp4,test/val_savanna/0/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +105,105,458,test/val_plains/0/000040.mp4,test/val_plains/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +106,106,371,test/val_iceplains/1/000049.mp4,test/val_iceplains/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +107,107,230,test/val_savanna/3/000045.mp4,test/val_savanna/3/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +108,108,372,test/val_iceplains/2/000049.mp4,test/val_iceplains/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +109,109,223,test/val_savanna/1/000042.mp4,test/val_savanna/1/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +110,110,68,test/val_desert/0/000026.mp4,test/val_desert/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +111,111,461,test/val_plains/2/000042.mp4,test/val_plains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +112,112,151,test/val_savanna/1/000009.mp4,test/val_savanna/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +113,113,72,test/val_desert/3/000027.mp4,test/val_desert/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +114,114,427,test/val_plains/0/000024.mp4,test/val_plains/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +115,115,317,test/val_iceplains/3/000029.mp4,test/val_iceplains/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +116,116,249,test/val_iceplains/2/000001.mp4,test/val_iceplains/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +117,117,364,test/val_iceplains/2/000046.mp4,test/val_iceplains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +118,118,96,test/val_desert/2/000037.mp4,test/val_desert/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +119,119,190,test/val_savanna/3/000029.mp4,test/val_savanna/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +120,120,227,test/val_savanna/2/000044.mp4,test/val_savanna/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +121,121,244,test/val_iceplains/0/000000.mp4,test/val_iceplains/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +122,122,38,test/val_desert/3/000015.mp4,test/val_desert/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +123,123,291,test/val_iceplains/2/000017.mp4,test/val_iceplains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +124,124,85,test/val_desert/2/000032.mp4,test/val_desert/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +125,125,358,test/val_iceplains/0/000044.mp4,test/val_iceplains/0/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +126,126,50,test/val_desert/1/000020.mp4,test/val_desert/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +127,127,207,test/val_savanna/0/000036.mp4,test/val_savanna/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +128,128,330,test/val_iceplains/2/000034.mp4,test/val_iceplains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +129,129,132,test/val_savanna/0/000001.mp4,test/val_savanna/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +130,130,423,test/val_plains/3/000021.mp4,test/val_plains/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +131,131,66,test/val_desert/2/000025.mp4,test/val_desert/2/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +132,132,153,test/val_savanna/3/000009.mp4,test/val_savanna/3/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +133,133,382,test/val_plains/0/000004.mp4,test/val_plains/0/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +134,134,143,test/val_savanna/2/000005.mp4,test/val_savanna/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +135,135,348,test/val_iceplains/1/000040.mp4,test/val_iceplains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +136,136,384,test/val_plains/2/000005.mp4,test/val_plains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +137,137,30,test/val_desert/0/000012.mp4,test/val_desert/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +138,138,203,test/val_savanna/2/000034.mp4,test/val_savanna/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +139,139,179,test/val_savanna/0/000022.mp4,test/val_savanna/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +140,140,253,test/val_iceplains/2/000003.mp4,test/val_iceplains/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +141,141,289,test/val_iceplains/2/000016.mp4,test/val_iceplains/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +142,142,422,test/val_plains/2/000020.mp4,test/val_plains/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +143,143,20,test/val_desert/0/000008.mp4,test/val_desert/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +144,144,130,test/val_savanna/0/000000.mp4,test/val_savanna/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +145,145,296,test/val_iceplains/1/000020.mp4,test/val_iceplains/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +146,146,278,test/val_iceplains/3/000013.mp4,test/val_iceplains/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +147,147,269,test/val_iceplains/0/000009.mp4,test/val_iceplains/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +148,148,314,test/val_iceplains/0/000028.mp4,test/val_iceplains/0/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +149,149,281,test/val_iceplains/0/000014.mp4,test/val_iceplains/0/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +150,150,149,test/val_savanna/0/000008.mp4,test/val_savanna/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +151,151,184,test/val_savanna/0/000024.mp4,test/val_savanna/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +152,152,407,test/val_plains/1/000014.mp4,test/val_plains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +153,153,304,test/val_iceplains/2/000023.mp4,test/val_iceplains/2/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +154,154,208,test/val_savanna/3/000037.mp4,test/val_savanna/3/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +155,155,182,test/val_savanna/3/000023.mp4,test/val_savanna/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +156,156,394,test/val_plains/1/000009.mp4,test/val_plains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +157,157,14,test/val_desert/0/000006.mp4,test/val_desert/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +158,158,55,test/val_desert/3/000021.mp4,test/val_desert/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +159,159,75,test/val_desert/2/000028.mp4,test/val_desert/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +160,160,319,test/val_iceplains/1/000031.mp4,test/val_iceplains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +161,161,39,test/val_desert/1/000016.mp4,test/val_desert/1/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +162,162,418,test/val_plains/1/000019.mp4,test/val_plains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +163,163,389,test/val_plains/3/000007.mp4,test/val_plains/3/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +164,164,409,test/val_plains/2/000014.mp4,test/val_plains/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +165,165,344,test/val_iceplains/2/000038.mp4,test/val_iceplains/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +166,166,396,test/val_plains/2/000009.mp4,test/val_plains/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +167,167,67,test/val_desert/1/000026.mp4,test/val_desert/1/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +168,168,99,test/val_desert/3/000038.mp4,test/val_desert/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +169,169,288,test/val_iceplains/3/000016.mp4,test/val_iceplains/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +170,170,97,test/val_desert/1/000038.mp4,test/val_desert/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +171,171,178,test/val_savanna/2/000021.mp4,test/val_savanna/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +172,172,264,test/val_iceplains/1/000007.mp4,test/val_iceplains/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +173,173,266,test/val_iceplains/0/000008.mp4,test/val_iceplains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +174,174,303,test/val_iceplains/1/000023.mp4,test/val_iceplains/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +175,175,442,test/val_plains/0/000031.mp4,test/val_plains/0/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +176,176,233,test/val_savanna/1/000047.mp4,test/val_savanna/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +177,177,240,test/val_savanna/1/000049.mp4,test/val_savanna/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +178,178,42,test/val_desert/0/000017.mp4,test/val_desert/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +179,179,369,test/val_iceplains/3/000048.mp4,test/val_iceplains/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +180,180,398,test/val_plains/3/000010.mp4,test/val_plains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +181,181,435,test/val_plains/2/000027.mp4,test/val_plains/2/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +182,182,242,test/val_savanna/2/000049.mp4,test/val_savanna/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +183,183,361,test/val_iceplains/2/000045.mp4,test/val_iceplains/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +184,184,470,test/val_plains/2/000047.mp4,test/val_plains/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +185,185,356,test/val_iceplains/2/000042.mp4,test/val_iceplains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +186,186,468,test/val_plains/2/000046.mp4,test/val_plains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +187,187,113,test/val_desert/0/000043.mp4,test/val_desert/0/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +188,188,19,test/val_desert/1/000008.mp4,test/val_desert/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +189,189,199,test/val_savanna/0/000033.mp4,test/val_savanna/0/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +190,190,118,test/val_desert/2/000045.mp4,test/val_desert/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +191,191,433,test/val_plains/0/000027.mp4,test/val_plains/0/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +192,192,328,test/val_iceplains/0/000034.mp4,test/val_iceplains/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +193,193,213,test/val_savanna/0/000039.mp4,test/val_savanna/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +194,194,339,test/val_iceplains/0/000037.mp4,test/val_iceplains/0/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +195,195,466,test/val_plains/1/000046.mp4,test/val_plains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +196,196,355,test/val_iceplains/3/000042.mp4,test/val_iceplains/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +197,197,336,test/val_iceplains/3/000036.mp4,test/val_iceplains/3/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +198,198,420,test/val_plains/0/000020.mp4,test/val_plains/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +199,199,410,test/val_plains/1/000015.mp4,test/val_plains/1/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +200,200,27,test/val_desert/1/000011.mp4,test/val_desert/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +201,201,86,test/val_desert/1/000033.mp4,test/val_desert/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +202,202,210,test/val_savanna/3/000038.mp4,test/val_savanna/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +203,203,103,test/val_desert/3/000039.mp4,test/val_desert/3/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +204,204,245,test/val_iceplains/3/000000.mp4,test/val_iceplains/3/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +205,205,25,test/val_desert/2/000009.mp4,test/val_desert/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +206,206,340,test/val_iceplains/2/000037.mp4,test/val_iceplains/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +207,207,167,test/val_savanna/0/000016.mp4,test/val_savanna/0/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +208,208,424,test/val_plains/0/000022.mp4,test/val_plains/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +209,209,448,test/val_plains/2/000033.mp4,test/val_plains/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +210,210,386,test/val_plains/0/000006.mp4,test/val_plains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +211,211,392,test/val_plains/3/000008.mp4,test/val_plains/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +212,212,105,test/val_desert/1/000040.mp4,test/val_desert/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +213,213,267,test/val_iceplains/2/000008.mp4,test/val_iceplains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +214,214,90,test/val_desert/0/000034.mp4,test/val_desert/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +215,215,426,test/val_plains/0/000023.mp4,test/val_plains/0/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +216,216,168,test/val_savanna/3/000016.mp4,test/val_savanna/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +217,217,218,test/val_savanna/3/000040.mp4,test/val_savanna/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +218,218,445,test/val_plains/0/000032.mp4,test/val_plains/0/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +219,219,48,test/val_desert/3/000019.mp4,test/val_desert/3/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +220,220,104,test/val_desert/2/000039.mp4,test/val_desert/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +221,221,206,test/val_savanna/2/000035.mp4,test/val_savanna/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +222,222,228,test/val_savanna/1/000045.mp4,test/val_savanna/1/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +223,223,200,test/val_savanna/2/000033.mp4,test/val_savanna/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +224,224,62,test/val_desert/2/000024.mp4,test/val_desert/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +225,225,121,test/val_desert/1/000047.mp4,test/val_desert/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +226,226,147,test/val_savanna/0/000007.mp4,test/val_savanna/0/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +227,227,146,test/val_savanna/2/000006.mp4,test/val_savanna/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +228,228,225,test/val_savanna/1/000043.mp4,test/val_savanna/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +229,229,313,test/val_iceplains/3/000027.mp4,test/val_iceplains/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +230,230,354,test/val_iceplains/0/000042.mp4,test/val_iceplains/0/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +231,231,367,test/val_iceplains/1/000048.mp4,test/val_iceplains/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +232,232,297,test/val_iceplains/3/000020.mp4,test/val_iceplains/3/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +233,233,124,test/val_desert/1/000048.mp4,test/val_desert/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +234,234,171,test/val_savanna/2/000017.mp4,test/val_savanna/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +235,235,342,test/val_iceplains/0/000038.mp4,test/val_iceplains/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +236,236,60,test/val_desert/3/000023.mp4,test/val_desert/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +237,237,131,test/val_savanna/1/000001.mp4,test/val_savanna/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +238,238,365,test/val_iceplains/1/000047.mp4,test/val_iceplains/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +239,239,315,test/val_iceplains/2/000028.mp4,test/val_iceplains/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +240,240,220,test/val_savanna/1/000041.mp4,test/val_savanna/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +241,241,8,test/val_desert/1/000003.mp4,test/val_desert/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +242,242,261,test/val_iceplains/0/000006.mp4,test/val_iceplains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +243,243,17,test/val_desert/1/000007.mp4,test/val_desert/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +244,244,462,test/val_plains/1/000043.mp4,test/val_plains/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +245,245,454,test/val_plains/3/000038.mp4,test/val_plains/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +246,246,351,test/val_iceplains/1/000041.mp4,test/val_iceplains/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +247,247,383,test/val_plains/3/000005.mp4,test/val_plains/3/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +248,248,18,test/val_desert/2/000007.mp4,test/val_desert/2/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +249,249,457,test/val_plains/1/000040.mp4,test/val_plains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +250,250,326,test/val_iceplains/3/000033.mp4,test/val_iceplains/3/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +251,251,254,test/val_iceplains/1/000004.mp4,test/val_iceplains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +252,252,375,test/val_plains/1/000002.mp4,test/val_plains/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +253,253,141,test/val_savanna/0/000005.mp4,test/val_savanna/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +254,254,306,test/val_iceplains/2/000024.mp4,test/val_iceplains/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +255,255,10,test/val_desert/2/000003.mp4,test/val_desert/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +256,256,173,test/val_savanna/0/000020.mp4,test/val_savanna/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +257,257,320,test/val_iceplains/1/000032.mp4,test/val_iceplains/1/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +258,258,115,test/val_desert/1/000044.mp4,test/val_desert/1/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +259,259,341,test/val_iceplains/1/000038.mp4,test/val_iceplains/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +260,260,61,test/val_desert/3/000024.mp4,test/val_desert/3/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +261,261,231,test/val_savanna/1/000046.mp4,test/val_savanna/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +262,262,224,test/val_savanna/3/000042.mp4,test/val_savanna/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +263,263,292,test/val_iceplains/3/000018.mp4,test/val_iceplains/3/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +264,264,334,test/val_iceplains/1/000036.mp4,test/val_iceplains/1/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +265,265,159,test/val_savanna/0/000013.mp4,test/val_savanna/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +266,266,102,test/val_desert/0/000039.mp4,test/val_desert/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +267,267,362,test/val_iceplains/1/000046.mp4,test/val_iceplains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +268,268,322,test/val_iceplains/3/000032.mp4,test/val_iceplains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +269,269,381,test/val_plains/1/000004.mp4,test/val_plains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +270,270,84,test/val_desert/3/000032.mp4,test/val_desert/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +271,271,243,test/val_iceplains/1/000000.mp4,test/val_iceplains/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +272,272,65,test/val_desert/3/000025.mp4,test/val_desert/3/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +273,273,64,test/val_desert/0/000025.mp4,test/val_desert/0/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +274,274,272,test/val_iceplains/2/000010.mp4,test/val_iceplains/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +275,275,399,test/val_plains/0/000011.mp4,test/val_plains/0/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +276,276,248,test/val_iceplains/3/000001.mp4,test/val_iceplains/3/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +277,277,43,test/val_desert/3/000017.mp4,test/val_desert/3/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +278,278,393,test/val_plains/2/000008.mp4,test/val_plains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +279,279,123,test/val_desert/2/000047.mp4,test/val_desert/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +280,280,29,test/val_desert/2/000011.mp4,test/val_desert/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +281,281,122,test/val_desert/0/000047.mp4,test/val_desert/0/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +282,282,157,test/val_savanna/3/000011.mp4,test/val_savanna/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +283,283,333,test/val_iceplains/2/000035.mp4,test/val_iceplains/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +284,284,307,test/val_iceplains/1/000025.mp4,test/val_iceplains/1/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +285,285,9,test/val_desert/0/000003.mp4,test/val_desert/0/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +286,286,250,test/val_iceplains/2/000002.mp4,test/val_iceplains/2/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +287,287,165,test/val_savanna/3/000015.mp4,test/val_savanna/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +288,288,0,test/val_desert/1/000000.mp4,test/val_desert/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +289,289,271,test/val_iceplains/3/000010.mp4,test/val_iceplains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +290,290,276,test/val_iceplains/1/000013.mp4,test/val_iceplains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +291,291,82,test/val_desert/1/000031.mp4,test/val_desert/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +292,292,191,test/val_savanna/2/000029.mp4,test/val_savanna/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +293,293,374,test/val_plains/1/000001.mp4,test/val_plains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +294,294,446,test/val_plains/3/000032.mp4,test/val_plains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +295,295,378,test/val_plains/1/000003.mp4,test/val_plains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +296,296,152,test/val_savanna/0/000009.mp4,test/val_savanna/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +297,297,54,test/val_desert/0/000021.mp4,test/val_desert/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +298,298,259,test/val_iceplains/2/000005.mp4,test/val_iceplains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +299,299,91,test/val_desert/2/000034.mp4,test/val_desert/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.sha256 b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..065350069e5b0f46e33cd51ad63f60aa956822fd --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_noise_off_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.sha256 @@ -0,0 +1 @@ +ef4e4de9d3975d8a7e724bb609e57f20186cd1f7285457c00fb85c1aca973b99 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/evaluation.patch b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/evaluation.patch new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..3cfd08bc49f4159d2bd5cace6409ac92956d78e1 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.csv @@ -0,0 +1,51 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.sha256 b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..67b0dc2042229c0768d324282215bdf64bebe28f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/fid50_manifest.sha256 @@ -0,0 +1 @@ +f9536911e434610cce23306ff99575db53cd7fc29e3df841da4b6631bf418634 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/git_provenance.json b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/git_provenance.json new file mode 100644 index 0000000000000000000000000000000000000000..fabe691fc08806f957f2445d4a52b2f38d7f07d8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/git_provenance.json @@ -0,0 +1,5 @@ +{ + "branch": "dememwm-unified-sourcefree-age-balanced", + "head": "732555281e86c4fdbb71c98a000d648c3edc3109", + "status": "?? results/\n?? train_dememwm_full_berzelius_gate_off_h200.sh\n?? train_dememwm_separate_mem_gatefix_4layers512_valid500_berzelius_h200.sh" +} diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..dae70b4071c2d2a834268c2a26f061d308f11f2c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.csv @@ -0,0 +1,301 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +50,50,212,test/val_savanna/1/000039.mp4,test/val_savanna/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +51,51,4,test/val_desert/0/000001.mp4,test/val_desert/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +52,52,7,test/val_desert/0/000002.mp4,test/val_desert/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +53,53,310,test/val_iceplains/0/000026.mp4,test/val_iceplains/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +54,54,284,test/val_iceplains/0/000015.mp4,test/val_iceplains/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +55,55,109,test/val_desert/3/000041.mp4,test/val_desert/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +56,56,45,test/val_desert/0/000018.mp4,test/val_desert/0/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +57,57,93,test/val_desert/0/000035.mp4,test/val_desert/0/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +58,58,400,test/val_plains/3/000011.mp4,test/val_plains/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +59,59,300,test/val_iceplains/2/000021.mp4,test/val_iceplains/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +60,60,226,test/val_savanna/2/000043.mp4,test/val_savanna/2/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +61,61,417,test/val_plains/2/000018.mp4,test/val_plains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +62,62,120,test/val_desert/3/000046.mp4,test/val_desert/3/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +63,63,106,test/val_desert/3/000040.mp4,test/val_desert/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +64,64,345,test/val_iceplains/1/000039.mp4,test/val_iceplains/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +65,65,169,test/val_savanna/2/000016.mp4,test/val_savanna/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +66,66,312,test/val_iceplains/1/000027.mp4,test/val_iceplains/1/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +67,67,368,test/val_iceplains/0/000048.mp4,test/val_iceplains/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +68,68,449,test/val_plains/1/000034.mp4,test/val_plains/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +69,69,11,test/val_desert/1/000004.mp4,test/val_desert/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +70,70,6,test/val_desert/2/000001.mp4,test/val_desert/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +71,71,257,test/val_iceplains/0/000005.mp4,test/val_iceplains/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +72,72,262,test/val_iceplains/3/000006.mp4,test/val_iceplains/3/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +73,73,164,test/val_savanna/0/000015.mp4,test/val_savanna/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +74,74,337,test/val_iceplains/2/000036.mp4,test/val_iceplains/2/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +75,75,95,test/val_desert/0/000036.mp4,test/val_desert/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +76,76,277,test/val_iceplains/0/000013.mp4,test/val_iceplains/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +77,77,26,test/val_desert/1/000010.mp4,test/val_desert/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +78,78,77,test/val_desert/0/000029.mp4,test/val_desert/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +79,79,294,test/val_iceplains/1/000019.mp4,test/val_iceplains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +80,80,34,test/val_desert/1/000014.mp4,test/val_desert/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +81,81,188,test/val_savanna/1/000029.mp4,test/val_savanna/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +82,82,170,test/val_savanna/0/000017.mp4,test/val_savanna/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +83,83,415,test/val_plains/2/000017.mp4,test/val_plains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +84,84,145,test/val_savanna/0/000006.mp4,test/val_savanna/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +85,85,133,test/val_savanna/1/000002.mp4,test/val_savanna/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +86,86,347,test/val_iceplains/2/000039.mp4,test/val_iceplains/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +87,87,209,test/val_savanna/0/000038.mp4,test/val_savanna/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +88,88,404,test/val_plains/1/000013.mp4,test/val_plains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +89,89,211,test/val_savanna/2/000038.mp4,test/val_savanna/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +90,90,464,test/val_plains/2/000044.mp4,test/val_plains/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +91,91,162,test/val_savanna/1/000014.mp4,test/val_savanna/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +92,92,419,test/val_plains/2/000019.mp4,test/val_plains/2/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +93,93,127,test/val_desert/3/000049.mp4,test/val_desert/3/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +94,94,370,test/val_iceplains/2/000048.mp4,test/val_iceplains/2/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +95,95,293,test/val_iceplains/2/000018.mp4,test/val_iceplains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +96,96,251,test/val_iceplains/1/000003.mp4,test/val_iceplains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +97,97,441,test/val_plains/1/000031.mp4,test/val_plains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +98,98,69,test/val_desert/2/000026.mp4,test/val_desert/2/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +99,99,286,test/val_iceplains/2/000015.mp4,test/val_iceplains/2/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +100,100,438,test/val_plains/0/000029.mp4,test/val_plains/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +101,101,376,test/val_plains/0/000002.mp4,test/val_plains/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +102,102,163,test/val_savanna/2/000014.mp4,test/val_savanna/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +103,103,31,test/val_desert/1/000013.mp4,test/val_desert/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +104,104,241,test/val_savanna/0/000049.mp4,test/val_savanna/0/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +105,105,458,test/val_plains/0/000040.mp4,test/val_plains/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +106,106,371,test/val_iceplains/1/000049.mp4,test/val_iceplains/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +107,107,230,test/val_savanna/3/000045.mp4,test/val_savanna/3/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +108,108,372,test/val_iceplains/2/000049.mp4,test/val_iceplains/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +109,109,223,test/val_savanna/1/000042.mp4,test/val_savanna/1/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +110,110,68,test/val_desert/0/000026.mp4,test/val_desert/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +111,111,461,test/val_plains/2/000042.mp4,test/val_plains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +112,112,151,test/val_savanna/1/000009.mp4,test/val_savanna/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +113,113,72,test/val_desert/3/000027.mp4,test/val_desert/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +114,114,427,test/val_plains/0/000024.mp4,test/val_plains/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +115,115,317,test/val_iceplains/3/000029.mp4,test/val_iceplains/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +116,116,249,test/val_iceplains/2/000001.mp4,test/val_iceplains/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +117,117,364,test/val_iceplains/2/000046.mp4,test/val_iceplains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +118,118,96,test/val_desert/2/000037.mp4,test/val_desert/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +119,119,190,test/val_savanna/3/000029.mp4,test/val_savanna/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +120,120,227,test/val_savanna/2/000044.mp4,test/val_savanna/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +121,121,244,test/val_iceplains/0/000000.mp4,test/val_iceplains/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +122,122,38,test/val_desert/3/000015.mp4,test/val_desert/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +123,123,291,test/val_iceplains/2/000017.mp4,test/val_iceplains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +124,124,85,test/val_desert/2/000032.mp4,test/val_desert/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +125,125,358,test/val_iceplains/0/000044.mp4,test/val_iceplains/0/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +126,126,50,test/val_desert/1/000020.mp4,test/val_desert/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +127,127,207,test/val_savanna/0/000036.mp4,test/val_savanna/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +128,128,330,test/val_iceplains/2/000034.mp4,test/val_iceplains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +129,129,132,test/val_savanna/0/000001.mp4,test/val_savanna/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +130,130,423,test/val_plains/3/000021.mp4,test/val_plains/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +131,131,66,test/val_desert/2/000025.mp4,test/val_desert/2/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +132,132,153,test/val_savanna/3/000009.mp4,test/val_savanna/3/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +133,133,382,test/val_plains/0/000004.mp4,test/val_plains/0/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +134,134,143,test/val_savanna/2/000005.mp4,test/val_savanna/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +135,135,348,test/val_iceplains/1/000040.mp4,test/val_iceplains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +136,136,384,test/val_plains/2/000005.mp4,test/val_plains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +137,137,30,test/val_desert/0/000012.mp4,test/val_desert/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +138,138,203,test/val_savanna/2/000034.mp4,test/val_savanna/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +139,139,179,test/val_savanna/0/000022.mp4,test/val_savanna/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +140,140,253,test/val_iceplains/2/000003.mp4,test/val_iceplains/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +141,141,289,test/val_iceplains/2/000016.mp4,test/val_iceplains/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +142,142,422,test/val_plains/2/000020.mp4,test/val_plains/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +143,143,20,test/val_desert/0/000008.mp4,test/val_desert/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +144,144,130,test/val_savanna/0/000000.mp4,test/val_savanna/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +145,145,296,test/val_iceplains/1/000020.mp4,test/val_iceplains/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +146,146,278,test/val_iceplains/3/000013.mp4,test/val_iceplains/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +147,147,269,test/val_iceplains/0/000009.mp4,test/val_iceplains/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +148,148,314,test/val_iceplains/0/000028.mp4,test/val_iceplains/0/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +149,149,281,test/val_iceplains/0/000014.mp4,test/val_iceplains/0/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +150,150,149,test/val_savanna/0/000008.mp4,test/val_savanna/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +151,151,184,test/val_savanna/0/000024.mp4,test/val_savanna/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +152,152,407,test/val_plains/1/000014.mp4,test/val_plains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +153,153,304,test/val_iceplains/2/000023.mp4,test/val_iceplains/2/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +154,154,208,test/val_savanna/3/000037.mp4,test/val_savanna/3/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +155,155,182,test/val_savanna/3/000023.mp4,test/val_savanna/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +156,156,394,test/val_plains/1/000009.mp4,test/val_plains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +157,157,14,test/val_desert/0/000006.mp4,test/val_desert/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +158,158,55,test/val_desert/3/000021.mp4,test/val_desert/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +159,159,75,test/val_desert/2/000028.mp4,test/val_desert/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +160,160,319,test/val_iceplains/1/000031.mp4,test/val_iceplains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +161,161,39,test/val_desert/1/000016.mp4,test/val_desert/1/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +162,162,418,test/val_plains/1/000019.mp4,test/val_plains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +163,163,389,test/val_plains/3/000007.mp4,test/val_plains/3/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +164,164,409,test/val_plains/2/000014.mp4,test/val_plains/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +165,165,344,test/val_iceplains/2/000038.mp4,test/val_iceplains/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +166,166,396,test/val_plains/2/000009.mp4,test/val_plains/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +167,167,67,test/val_desert/1/000026.mp4,test/val_desert/1/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +168,168,99,test/val_desert/3/000038.mp4,test/val_desert/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +169,169,288,test/val_iceplains/3/000016.mp4,test/val_iceplains/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +170,170,97,test/val_desert/1/000038.mp4,test/val_desert/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +171,171,178,test/val_savanna/2/000021.mp4,test/val_savanna/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +172,172,264,test/val_iceplains/1/000007.mp4,test/val_iceplains/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +173,173,266,test/val_iceplains/0/000008.mp4,test/val_iceplains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +174,174,303,test/val_iceplains/1/000023.mp4,test/val_iceplains/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +175,175,442,test/val_plains/0/000031.mp4,test/val_plains/0/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +176,176,233,test/val_savanna/1/000047.mp4,test/val_savanna/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +177,177,240,test/val_savanna/1/000049.mp4,test/val_savanna/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +178,178,42,test/val_desert/0/000017.mp4,test/val_desert/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +179,179,369,test/val_iceplains/3/000048.mp4,test/val_iceplains/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +180,180,398,test/val_plains/3/000010.mp4,test/val_plains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +181,181,435,test/val_plains/2/000027.mp4,test/val_plains/2/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +182,182,242,test/val_savanna/2/000049.mp4,test/val_savanna/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +183,183,361,test/val_iceplains/2/000045.mp4,test/val_iceplains/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +184,184,470,test/val_plains/2/000047.mp4,test/val_plains/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +185,185,356,test/val_iceplains/2/000042.mp4,test/val_iceplains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +186,186,468,test/val_plains/2/000046.mp4,test/val_plains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +187,187,113,test/val_desert/0/000043.mp4,test/val_desert/0/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +188,188,19,test/val_desert/1/000008.mp4,test/val_desert/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +189,189,199,test/val_savanna/0/000033.mp4,test/val_savanna/0/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +190,190,118,test/val_desert/2/000045.mp4,test/val_desert/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +191,191,433,test/val_plains/0/000027.mp4,test/val_plains/0/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +192,192,328,test/val_iceplains/0/000034.mp4,test/val_iceplains/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +193,193,213,test/val_savanna/0/000039.mp4,test/val_savanna/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +194,194,339,test/val_iceplains/0/000037.mp4,test/val_iceplains/0/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +195,195,466,test/val_plains/1/000046.mp4,test/val_plains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +196,196,355,test/val_iceplains/3/000042.mp4,test/val_iceplains/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +197,197,336,test/val_iceplains/3/000036.mp4,test/val_iceplains/3/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +198,198,420,test/val_plains/0/000020.mp4,test/val_plains/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +199,199,410,test/val_plains/1/000015.mp4,test/val_plains/1/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +200,200,27,test/val_desert/1/000011.mp4,test/val_desert/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +201,201,86,test/val_desert/1/000033.mp4,test/val_desert/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +202,202,210,test/val_savanna/3/000038.mp4,test/val_savanna/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +203,203,103,test/val_desert/3/000039.mp4,test/val_desert/3/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +204,204,245,test/val_iceplains/3/000000.mp4,test/val_iceplains/3/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +205,205,25,test/val_desert/2/000009.mp4,test/val_desert/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +206,206,340,test/val_iceplains/2/000037.mp4,test/val_iceplains/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +207,207,167,test/val_savanna/0/000016.mp4,test/val_savanna/0/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +208,208,424,test/val_plains/0/000022.mp4,test/val_plains/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +209,209,448,test/val_plains/2/000033.mp4,test/val_plains/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +210,210,386,test/val_plains/0/000006.mp4,test/val_plains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +211,211,392,test/val_plains/3/000008.mp4,test/val_plains/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +212,212,105,test/val_desert/1/000040.mp4,test/val_desert/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +213,213,267,test/val_iceplains/2/000008.mp4,test/val_iceplains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +214,214,90,test/val_desert/0/000034.mp4,test/val_desert/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +215,215,426,test/val_plains/0/000023.mp4,test/val_plains/0/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +216,216,168,test/val_savanna/3/000016.mp4,test/val_savanna/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +217,217,218,test/val_savanna/3/000040.mp4,test/val_savanna/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +218,218,445,test/val_plains/0/000032.mp4,test/val_plains/0/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +219,219,48,test/val_desert/3/000019.mp4,test/val_desert/3/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +220,220,104,test/val_desert/2/000039.mp4,test/val_desert/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +221,221,206,test/val_savanna/2/000035.mp4,test/val_savanna/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +222,222,228,test/val_savanna/1/000045.mp4,test/val_savanna/1/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +223,223,200,test/val_savanna/2/000033.mp4,test/val_savanna/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +224,224,62,test/val_desert/2/000024.mp4,test/val_desert/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +225,225,121,test/val_desert/1/000047.mp4,test/val_desert/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +226,226,147,test/val_savanna/0/000007.mp4,test/val_savanna/0/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +227,227,146,test/val_savanna/2/000006.mp4,test/val_savanna/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +228,228,225,test/val_savanna/1/000043.mp4,test/val_savanna/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +229,229,313,test/val_iceplains/3/000027.mp4,test/val_iceplains/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +230,230,354,test/val_iceplains/0/000042.mp4,test/val_iceplains/0/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +231,231,367,test/val_iceplains/1/000048.mp4,test/val_iceplains/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +232,232,297,test/val_iceplains/3/000020.mp4,test/val_iceplains/3/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +233,233,124,test/val_desert/1/000048.mp4,test/val_desert/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +234,234,171,test/val_savanna/2/000017.mp4,test/val_savanna/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +235,235,342,test/val_iceplains/0/000038.mp4,test/val_iceplains/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +236,236,60,test/val_desert/3/000023.mp4,test/val_desert/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +237,237,131,test/val_savanna/1/000001.mp4,test/val_savanna/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +238,238,365,test/val_iceplains/1/000047.mp4,test/val_iceplains/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +239,239,315,test/val_iceplains/2/000028.mp4,test/val_iceplains/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +240,240,220,test/val_savanna/1/000041.mp4,test/val_savanna/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +241,241,8,test/val_desert/1/000003.mp4,test/val_desert/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +242,242,261,test/val_iceplains/0/000006.mp4,test/val_iceplains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +243,243,17,test/val_desert/1/000007.mp4,test/val_desert/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +244,244,462,test/val_plains/1/000043.mp4,test/val_plains/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +245,245,454,test/val_plains/3/000038.mp4,test/val_plains/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +246,246,351,test/val_iceplains/1/000041.mp4,test/val_iceplains/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +247,247,383,test/val_plains/3/000005.mp4,test/val_plains/3/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +248,248,18,test/val_desert/2/000007.mp4,test/val_desert/2/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +249,249,457,test/val_plains/1/000040.mp4,test/val_plains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +250,250,326,test/val_iceplains/3/000033.mp4,test/val_iceplains/3/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +251,251,254,test/val_iceplains/1/000004.mp4,test/val_iceplains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +252,252,375,test/val_plains/1/000002.mp4,test/val_plains/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +253,253,141,test/val_savanna/0/000005.mp4,test/val_savanna/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +254,254,306,test/val_iceplains/2/000024.mp4,test/val_iceplains/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +255,255,10,test/val_desert/2/000003.mp4,test/val_desert/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +256,256,173,test/val_savanna/0/000020.mp4,test/val_savanna/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +257,257,320,test/val_iceplains/1/000032.mp4,test/val_iceplains/1/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +258,258,115,test/val_desert/1/000044.mp4,test/val_desert/1/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +259,259,341,test/val_iceplains/1/000038.mp4,test/val_iceplains/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +260,260,61,test/val_desert/3/000024.mp4,test/val_desert/3/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +261,261,231,test/val_savanna/1/000046.mp4,test/val_savanna/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +262,262,224,test/val_savanna/3/000042.mp4,test/val_savanna/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +263,263,292,test/val_iceplains/3/000018.mp4,test/val_iceplains/3/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +264,264,334,test/val_iceplains/1/000036.mp4,test/val_iceplains/1/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +265,265,159,test/val_savanna/0/000013.mp4,test/val_savanna/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +266,266,102,test/val_desert/0/000039.mp4,test/val_desert/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +267,267,362,test/val_iceplains/1/000046.mp4,test/val_iceplains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +268,268,322,test/val_iceplains/3/000032.mp4,test/val_iceplains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +269,269,381,test/val_plains/1/000004.mp4,test/val_plains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +270,270,84,test/val_desert/3/000032.mp4,test/val_desert/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +271,271,243,test/val_iceplains/1/000000.mp4,test/val_iceplains/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +272,272,65,test/val_desert/3/000025.mp4,test/val_desert/3/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +273,273,64,test/val_desert/0/000025.mp4,test/val_desert/0/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +274,274,272,test/val_iceplains/2/000010.mp4,test/val_iceplains/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +275,275,399,test/val_plains/0/000011.mp4,test/val_plains/0/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +276,276,248,test/val_iceplains/3/000001.mp4,test/val_iceplains/3/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +277,277,43,test/val_desert/3/000017.mp4,test/val_desert/3/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +278,278,393,test/val_plains/2/000008.mp4,test/val_plains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +279,279,123,test/val_desert/2/000047.mp4,test/val_desert/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +280,280,29,test/val_desert/2/000011.mp4,test/val_desert/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +281,281,122,test/val_desert/0/000047.mp4,test/val_desert/0/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +282,282,157,test/val_savanna/3/000011.mp4,test/val_savanna/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +283,283,333,test/val_iceplains/2/000035.mp4,test/val_iceplains/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +284,284,307,test/val_iceplains/1/000025.mp4,test/val_iceplains/1/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +285,285,9,test/val_desert/0/000003.mp4,test/val_desert/0/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +286,286,250,test/val_iceplains/2/000002.mp4,test/val_iceplains/2/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +287,287,165,test/val_savanna/3/000015.mp4,test/val_savanna/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +288,288,0,test/val_desert/1/000000.mp4,test/val_desert/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +289,289,271,test/val_iceplains/3/000010.mp4,test/val_iceplains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +290,290,276,test/val_iceplains/1/000013.mp4,test/val_iceplains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +291,291,82,test/val_desert/1/000031.mp4,test/val_desert/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +292,292,191,test/val_savanna/2/000029.mp4,test/val_savanna/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +293,293,374,test/val_plains/1/000001.mp4,test/val_plains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +294,294,446,test/val_plains/3/000032.mp4,test/val_plains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +295,295,378,test/val_plains/1/000003.mp4,test/val_plains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +296,296,152,test/val_savanna/0/000009.mp4,test/val_savanna/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +297,297,54,test/val_desert/0/000021.mp4,test/val_desert/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +298,298,259,test/val_iceplains/2/000005.mp4,test/val_iceplains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +299,299,91,test/val_desert/2/000034.mp4,test/val_desert/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.sha256 b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..065350069e5b0f46e33cd51ad63f60aa956822fd --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed42/test_manifest.sha256 @@ -0,0 +1 @@ +ef4e4de9d3975d8a7e724bb609e57f20186cd1f7285457c00fb85c1aca973b99 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/evaluation.patch b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/evaluation.patch new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..3cfd08bc49f4159d2bd5cace6409ac92956d78e1 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.csv @@ -0,0 +1,51 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.sha256 b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..67b0dc2042229c0768d324282215bdf64bebe28f --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/fid50_manifest.sha256 @@ -0,0 +1 @@ +f9536911e434610cce23306ff99575db53cd7fc29e3df841da4b6631bf418634 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/git_provenance.json b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/git_provenance.json new file mode 100644 index 0000000000000000000000000000000000000000..fabe691fc08806f957f2445d4a52b2f38d7f07d8 --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/git_provenance.json @@ -0,0 +1,5 @@ +{ + "branch": "dememwm-unified-sourcefree-age-balanced", + "head": "732555281e86c4fdbb71c98a000d648c3edc3109", + "status": "?? results/\n?? train_dememwm_full_berzelius_gate_off_h200.sh\n?? train_dememwm_separate_mem_gatefix_4layers512_valid500_berzelius_h200.sh" +} diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.csv b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.csv new file mode 100644 index 0000000000000000000000000000000000000000..dae70b4071c2d2a834268c2a26f061d308f11f2c --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.csv @@ -0,0 +1,301 @@ +manifest_index,global_test_index,dataset_index,relative_path,clip_id,source_offset,context_start,context_end,target_start,target_end,n_frames_valid,context_frames,target_frames,frame_skip,raw_frame_count,action_count,pose_count,latent_count +0,0,450,test/val_plains/2/000034.mp4,test/val_plains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +1,1,76,test/val_desert/1/000029.mp4,test/val_desert/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +2,2,88,test/val_desert/2/000033.mp4,test/val_desert/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +3,3,444,test/val_plains/2/000031.mp4,test/val_plains/2/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +4,4,201,test/val_savanna/1/000034.mp4,test/val_savanna/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +5,5,92,test/val_desert/1/000035.mp4,test/val_desert/1/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +6,6,268,test/val_iceplains/1/000009.mp4,test/val_iceplains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +7,7,156,test/val_savanna/1/000011.mp4,test/val_savanna/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +8,8,401,test/val_plains/2/000011.mp4,test/val_plains/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +9,9,100,test/val_desert/2/000038.mp4,test/val_desert/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +10,10,160,test/val_savanna/3/000013.mp4,test/val_savanna/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +11,11,238,test/val_savanna/3/000048.mp4,test/val_savanna/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +12,12,263,test/val_iceplains/2/000006.mp4,test/val_iceplains/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +13,13,196,test/val_savanna/3/000032.mp4,test/val_savanna/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +14,14,74,test/val_desert/3/000028.mp4,test/val_desert/3/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +15,15,247,test/val_iceplains/1/000001.mp4,test/val_iceplains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +16,16,21,test/val_desert/3/000008.mp4,test/val_desert/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +17,17,41,test/val_desert/2/000016.mp4,test/val_desert/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +18,18,221,test/val_savanna/3/000041.mp4,test/val_savanna/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +19,19,280,test/val_iceplains/1/000014.mp4,test/val_iceplains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +20,20,139,test/val_savanna/3/000004.mp4,test/val_savanna/3/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +21,21,217,test/val_savanna/0/000040.mp4,test/val_savanna/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +22,22,53,test/val_desert/1/000021.mp4,test/val_desert/1/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +23,23,402,test/val_plains/0/000012.mp4,test/val_plains/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +24,24,155,test/val_savanna/2/000010.mp4,test/val_savanna/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +25,25,439,test/val_plains/2/000029.mp4,test/val_plains/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +26,26,237,test/val_savanna/0/000048.mp4,test/val_savanna/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +27,27,144,test/val_savanna/1/000006.mp4,test/val_savanna/1/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +28,28,391,test/val_plains/0/000008.mp4,test/val_plains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +29,29,89,test/val_desert/1/000034.mp4,test/val_desert/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +30,30,246,test/val_iceplains/2/000000.mp4,test/val_iceplains/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +31,31,265,test/val_iceplains/1/000008.mp4,test/val_iceplains/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +32,32,318,test/val_iceplains/3/000030.mp4,test/val_iceplains/3/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +33,33,175,test/val_savanna/2/000020.mp4,test/val_savanna/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +34,34,154,test/val_savanna/1/000010.mp4,test/val_savanna/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +35,35,94,test/val_desert/3/000035.mp4,test/val_desert/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +36,36,436,test/val_plains/1/000028.mp4,test/val_plains/1/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +37,37,187,test/val_savanna/2/000028.mp4,test/val_savanna/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +38,38,192,test/val_savanna/2/000030.mp4,test/val_savanna/2/000030,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +39,39,323,test/val_iceplains/2/000032.mp4,test/val_iceplains/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +40,40,59,test/val_desert/1/000023.mp4,test/val_desert/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +41,41,177,test/val_savanna/0/000021.mp4,test/val_savanna/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +42,42,158,test/val_savanna/2/000011.mp4,test/val_savanna/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +43,43,198,test/val_savanna/1/000033.mp4,test/val_savanna/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +44,44,215,test/val_savanna/2/000039.mp4,test/val_savanna/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +45,45,37,test/val_desert/0/000015.mp4,test/val_desert/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +46,46,180,test/val_savanna/3/000022.mp4,test/val_savanna/3/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +47,47,2,test/val_desert/2/000000.mp4,test/val_desert/2/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +48,48,129,test/val_savanna/1/000000.mp4,test/val_savanna/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +49,49,452,test/val_plains/3/000035.mp4,test/val_plains/3/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +50,50,212,test/val_savanna/1/000039.mp4,test/val_savanna/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +51,51,4,test/val_desert/0/000001.mp4,test/val_desert/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +52,52,7,test/val_desert/0/000002.mp4,test/val_desert/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +53,53,310,test/val_iceplains/0/000026.mp4,test/val_iceplains/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +54,54,284,test/val_iceplains/0/000015.mp4,test/val_iceplains/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +55,55,109,test/val_desert/3/000041.mp4,test/val_desert/3/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +56,56,45,test/val_desert/0/000018.mp4,test/val_desert/0/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +57,57,93,test/val_desert/0/000035.mp4,test/val_desert/0/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +58,58,400,test/val_plains/3/000011.mp4,test/val_plains/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +59,59,300,test/val_iceplains/2/000021.mp4,test/val_iceplains/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +60,60,226,test/val_savanna/2/000043.mp4,test/val_savanna/2/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +61,61,417,test/val_plains/2/000018.mp4,test/val_plains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +62,62,120,test/val_desert/3/000046.mp4,test/val_desert/3/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +63,63,106,test/val_desert/3/000040.mp4,test/val_desert/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +64,64,345,test/val_iceplains/1/000039.mp4,test/val_iceplains/1/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +65,65,169,test/val_savanna/2/000016.mp4,test/val_savanna/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +66,66,312,test/val_iceplains/1/000027.mp4,test/val_iceplains/1/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +67,67,368,test/val_iceplains/0/000048.mp4,test/val_iceplains/0/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +68,68,449,test/val_plains/1/000034.mp4,test/val_plains/1/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +69,69,11,test/val_desert/1/000004.mp4,test/val_desert/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +70,70,6,test/val_desert/2/000001.mp4,test/val_desert/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +71,71,257,test/val_iceplains/0/000005.mp4,test/val_iceplains/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +72,72,262,test/val_iceplains/3/000006.mp4,test/val_iceplains/3/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +73,73,164,test/val_savanna/0/000015.mp4,test/val_savanna/0/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +74,74,337,test/val_iceplains/2/000036.mp4,test/val_iceplains/2/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +75,75,95,test/val_desert/0/000036.mp4,test/val_desert/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +76,76,277,test/val_iceplains/0/000013.mp4,test/val_iceplains/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +77,77,26,test/val_desert/1/000010.mp4,test/val_desert/1/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +78,78,77,test/val_desert/0/000029.mp4,test/val_desert/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +79,79,294,test/val_iceplains/1/000019.mp4,test/val_iceplains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +80,80,34,test/val_desert/1/000014.mp4,test/val_desert/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +81,81,188,test/val_savanna/1/000029.mp4,test/val_savanna/1/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +82,82,170,test/val_savanna/0/000017.mp4,test/val_savanna/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +83,83,415,test/val_plains/2/000017.mp4,test/val_plains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +84,84,145,test/val_savanna/0/000006.mp4,test/val_savanna/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +85,85,133,test/val_savanna/1/000002.mp4,test/val_savanna/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +86,86,347,test/val_iceplains/2/000039.mp4,test/val_iceplains/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +87,87,209,test/val_savanna/0/000038.mp4,test/val_savanna/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +88,88,404,test/val_plains/1/000013.mp4,test/val_plains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +89,89,211,test/val_savanna/2/000038.mp4,test/val_savanna/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +90,90,464,test/val_plains/2/000044.mp4,test/val_plains/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +91,91,162,test/val_savanna/1/000014.mp4,test/val_savanna/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +92,92,419,test/val_plains/2/000019.mp4,test/val_plains/2/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +93,93,127,test/val_desert/3/000049.mp4,test/val_desert/3/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +94,94,370,test/val_iceplains/2/000048.mp4,test/val_iceplains/2/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +95,95,293,test/val_iceplains/2/000018.mp4,test/val_iceplains/2/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +96,96,251,test/val_iceplains/1/000003.mp4,test/val_iceplains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +97,97,441,test/val_plains/1/000031.mp4,test/val_plains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +98,98,69,test/val_desert/2/000026.mp4,test/val_desert/2/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +99,99,286,test/val_iceplains/2/000015.mp4,test/val_iceplains/2/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +100,100,438,test/val_plains/0/000029.mp4,test/val_plains/0/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +101,101,376,test/val_plains/0/000002.mp4,test/val_plains/0/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +102,102,163,test/val_savanna/2/000014.mp4,test/val_savanna/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +103,103,31,test/val_desert/1/000013.mp4,test/val_desert/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +104,104,241,test/val_savanna/0/000049.mp4,test/val_savanna/0/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +105,105,458,test/val_plains/0/000040.mp4,test/val_plains/0/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +106,106,371,test/val_iceplains/1/000049.mp4,test/val_iceplains/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +107,107,230,test/val_savanna/3/000045.mp4,test/val_savanna/3/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +108,108,372,test/val_iceplains/2/000049.mp4,test/val_iceplains/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +109,109,223,test/val_savanna/1/000042.mp4,test/val_savanna/1/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +110,110,68,test/val_desert/0/000026.mp4,test/val_desert/0/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +111,111,461,test/val_plains/2/000042.mp4,test/val_plains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +112,112,151,test/val_savanna/1/000009.mp4,test/val_savanna/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +113,113,72,test/val_desert/3/000027.mp4,test/val_desert/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +114,114,427,test/val_plains/0/000024.mp4,test/val_plains/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +115,115,317,test/val_iceplains/3/000029.mp4,test/val_iceplains/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +116,116,249,test/val_iceplains/2/000001.mp4,test/val_iceplains/2/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +117,117,364,test/val_iceplains/2/000046.mp4,test/val_iceplains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +118,118,96,test/val_desert/2/000037.mp4,test/val_desert/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +119,119,190,test/val_savanna/3/000029.mp4,test/val_savanna/3/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +120,120,227,test/val_savanna/2/000044.mp4,test/val_savanna/2/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +121,121,244,test/val_iceplains/0/000000.mp4,test/val_iceplains/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +122,122,38,test/val_desert/3/000015.mp4,test/val_desert/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +123,123,291,test/val_iceplains/2/000017.mp4,test/val_iceplains/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +124,124,85,test/val_desert/2/000032.mp4,test/val_desert/2/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +125,125,358,test/val_iceplains/0/000044.mp4,test/val_iceplains/0/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +126,126,50,test/val_desert/1/000020.mp4,test/val_desert/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +127,127,207,test/val_savanna/0/000036.mp4,test/val_savanna/0/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +128,128,330,test/val_iceplains/2/000034.mp4,test/val_iceplains/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +129,129,132,test/val_savanna/0/000001.mp4,test/val_savanna/0/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +130,130,423,test/val_plains/3/000021.mp4,test/val_plains/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +131,131,66,test/val_desert/2/000025.mp4,test/val_desert/2/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +132,132,153,test/val_savanna/3/000009.mp4,test/val_savanna/3/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +133,133,382,test/val_plains/0/000004.mp4,test/val_plains/0/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +134,134,143,test/val_savanna/2/000005.mp4,test/val_savanna/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +135,135,348,test/val_iceplains/1/000040.mp4,test/val_iceplains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +136,136,384,test/val_plains/2/000005.mp4,test/val_plains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +137,137,30,test/val_desert/0/000012.mp4,test/val_desert/0/000012,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +138,138,203,test/val_savanna/2/000034.mp4,test/val_savanna/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +139,139,179,test/val_savanna/0/000022.mp4,test/val_savanna/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +140,140,253,test/val_iceplains/2/000003.mp4,test/val_iceplains/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +141,141,289,test/val_iceplains/2/000016.mp4,test/val_iceplains/2/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +142,142,422,test/val_plains/2/000020.mp4,test/val_plains/2/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +143,143,20,test/val_desert/0/000008.mp4,test/val_desert/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +144,144,130,test/val_savanna/0/000000.mp4,test/val_savanna/0/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +145,145,296,test/val_iceplains/1/000020.mp4,test/val_iceplains/1/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +146,146,278,test/val_iceplains/3/000013.mp4,test/val_iceplains/3/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +147,147,269,test/val_iceplains/0/000009.mp4,test/val_iceplains/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +148,148,314,test/val_iceplains/0/000028.mp4,test/val_iceplains/0/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +149,149,281,test/val_iceplains/0/000014.mp4,test/val_iceplains/0/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +150,150,149,test/val_savanna/0/000008.mp4,test/val_savanna/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +151,151,184,test/val_savanna/0/000024.mp4,test/val_savanna/0/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +152,152,407,test/val_plains/1/000014.mp4,test/val_plains/1/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +153,153,304,test/val_iceplains/2/000023.mp4,test/val_iceplains/2/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +154,154,208,test/val_savanna/3/000037.mp4,test/val_savanna/3/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +155,155,182,test/val_savanna/3/000023.mp4,test/val_savanna/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +156,156,394,test/val_plains/1/000009.mp4,test/val_plains/1/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +157,157,14,test/val_desert/0/000006.mp4,test/val_desert/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +158,158,55,test/val_desert/3/000021.mp4,test/val_desert/3/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +159,159,75,test/val_desert/2/000028.mp4,test/val_desert/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +160,160,319,test/val_iceplains/1/000031.mp4,test/val_iceplains/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +161,161,39,test/val_desert/1/000016.mp4,test/val_desert/1/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +162,162,418,test/val_plains/1/000019.mp4,test/val_plains/1/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +163,163,389,test/val_plains/3/000007.mp4,test/val_plains/3/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +164,164,409,test/val_plains/2/000014.mp4,test/val_plains/2/000014,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +165,165,344,test/val_iceplains/2/000038.mp4,test/val_iceplains/2/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +166,166,396,test/val_plains/2/000009.mp4,test/val_plains/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +167,167,67,test/val_desert/1/000026.mp4,test/val_desert/1/000026,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +168,168,99,test/val_desert/3/000038.mp4,test/val_desert/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +169,169,288,test/val_iceplains/3/000016.mp4,test/val_iceplains/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +170,170,97,test/val_desert/1/000038.mp4,test/val_desert/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +171,171,178,test/val_savanna/2/000021.mp4,test/val_savanna/2/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +172,172,264,test/val_iceplains/1/000007.mp4,test/val_iceplains/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +173,173,266,test/val_iceplains/0/000008.mp4,test/val_iceplains/0/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +174,174,303,test/val_iceplains/1/000023.mp4,test/val_iceplains/1/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +175,175,442,test/val_plains/0/000031.mp4,test/val_plains/0/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +176,176,233,test/val_savanna/1/000047.mp4,test/val_savanna/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +177,177,240,test/val_savanna/1/000049.mp4,test/val_savanna/1/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +178,178,42,test/val_desert/0/000017.mp4,test/val_desert/0/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +179,179,369,test/val_iceplains/3/000048.mp4,test/val_iceplains/3/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +180,180,398,test/val_plains/3/000010.mp4,test/val_plains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +181,181,435,test/val_plains/2/000027.mp4,test/val_plains/2/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +182,182,242,test/val_savanna/2/000049.mp4,test/val_savanna/2/000049,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +183,183,361,test/val_iceplains/2/000045.mp4,test/val_iceplains/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +184,184,470,test/val_plains/2/000047.mp4,test/val_plains/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +185,185,356,test/val_iceplains/2/000042.mp4,test/val_iceplains/2/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +186,186,468,test/val_plains/2/000046.mp4,test/val_plains/2/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +187,187,113,test/val_desert/0/000043.mp4,test/val_desert/0/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +188,188,19,test/val_desert/1/000008.mp4,test/val_desert/1/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +189,189,199,test/val_savanna/0/000033.mp4,test/val_savanna/0/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +190,190,118,test/val_desert/2/000045.mp4,test/val_desert/2/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +191,191,433,test/val_plains/0/000027.mp4,test/val_plains/0/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +192,192,328,test/val_iceplains/0/000034.mp4,test/val_iceplains/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +193,193,213,test/val_savanna/0/000039.mp4,test/val_savanna/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +194,194,339,test/val_iceplains/0/000037.mp4,test/val_iceplains/0/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +195,195,466,test/val_plains/1/000046.mp4,test/val_plains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +196,196,355,test/val_iceplains/3/000042.mp4,test/val_iceplains/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +197,197,336,test/val_iceplains/3/000036.mp4,test/val_iceplains/3/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +198,198,420,test/val_plains/0/000020.mp4,test/val_plains/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +199,199,410,test/val_plains/1/000015.mp4,test/val_plains/1/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +200,200,27,test/val_desert/1/000011.mp4,test/val_desert/1/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +201,201,86,test/val_desert/1/000033.mp4,test/val_desert/1/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +202,202,210,test/val_savanna/3/000038.mp4,test/val_savanna/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +203,203,103,test/val_desert/3/000039.mp4,test/val_desert/3/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +204,204,245,test/val_iceplains/3/000000.mp4,test/val_iceplains/3/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +205,205,25,test/val_desert/2/000009.mp4,test/val_desert/2/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +206,206,340,test/val_iceplains/2/000037.mp4,test/val_iceplains/2/000037,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +207,207,167,test/val_savanna/0/000016.mp4,test/val_savanna/0/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +208,208,424,test/val_plains/0/000022.mp4,test/val_plains/0/000022,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +209,209,448,test/val_plains/2/000033.mp4,test/val_plains/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +210,210,386,test/val_plains/0/000006.mp4,test/val_plains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +211,211,392,test/val_plains/3/000008.mp4,test/val_plains/3/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +212,212,105,test/val_desert/1/000040.mp4,test/val_desert/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +213,213,267,test/val_iceplains/2/000008.mp4,test/val_iceplains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +214,214,90,test/val_desert/0/000034.mp4,test/val_desert/0/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +215,215,426,test/val_plains/0/000023.mp4,test/val_plains/0/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +216,216,168,test/val_savanna/3/000016.mp4,test/val_savanna/3/000016,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +217,217,218,test/val_savanna/3/000040.mp4,test/val_savanna/3/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +218,218,445,test/val_plains/0/000032.mp4,test/val_plains/0/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +219,219,48,test/val_desert/3/000019.mp4,test/val_desert/3/000019,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +220,220,104,test/val_desert/2/000039.mp4,test/val_desert/2/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +221,221,206,test/val_savanna/2/000035.mp4,test/val_savanna/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +222,222,228,test/val_savanna/1/000045.mp4,test/val_savanna/1/000045,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +223,223,200,test/val_savanna/2/000033.mp4,test/val_savanna/2/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +224,224,62,test/val_desert/2/000024.mp4,test/val_desert/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +225,225,121,test/val_desert/1/000047.mp4,test/val_desert/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +226,226,147,test/val_savanna/0/000007.mp4,test/val_savanna/0/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +227,227,146,test/val_savanna/2/000006.mp4,test/val_savanna/2/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +228,228,225,test/val_savanna/1/000043.mp4,test/val_savanna/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +229,229,313,test/val_iceplains/3/000027.mp4,test/val_iceplains/3/000027,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +230,230,354,test/val_iceplains/0/000042.mp4,test/val_iceplains/0/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +231,231,367,test/val_iceplains/1/000048.mp4,test/val_iceplains/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +232,232,297,test/val_iceplains/3/000020.mp4,test/val_iceplains/3/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +233,233,124,test/val_desert/1/000048.mp4,test/val_desert/1/000048,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +234,234,171,test/val_savanna/2/000017.mp4,test/val_savanna/2/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +235,235,342,test/val_iceplains/0/000038.mp4,test/val_iceplains/0/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +236,236,60,test/val_desert/3/000023.mp4,test/val_desert/3/000023,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +237,237,131,test/val_savanna/1/000001.mp4,test/val_savanna/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +238,238,365,test/val_iceplains/1/000047.mp4,test/val_iceplains/1/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +239,239,315,test/val_iceplains/2/000028.mp4,test/val_iceplains/2/000028,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +240,240,220,test/val_savanna/1/000041.mp4,test/val_savanna/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +241,241,8,test/val_desert/1/000003.mp4,test/val_desert/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +242,242,261,test/val_iceplains/0/000006.mp4,test/val_iceplains/0/000006,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +243,243,17,test/val_desert/1/000007.mp4,test/val_desert/1/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +244,244,462,test/val_plains/1/000043.mp4,test/val_plains/1/000043,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +245,245,454,test/val_plains/3/000038.mp4,test/val_plains/3/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +246,246,351,test/val_iceplains/1/000041.mp4,test/val_iceplains/1/000041,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +247,247,383,test/val_plains/3/000005.mp4,test/val_plains/3/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +248,248,18,test/val_desert/2/000007.mp4,test/val_desert/2/000007,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +249,249,457,test/val_plains/1/000040.mp4,test/val_plains/1/000040,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +250,250,326,test/val_iceplains/3/000033.mp4,test/val_iceplains/3/000033,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +251,251,254,test/val_iceplains/1/000004.mp4,test/val_iceplains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +252,252,375,test/val_plains/1/000002.mp4,test/val_plains/1/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +253,253,141,test/val_savanna/0/000005.mp4,test/val_savanna/0/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +254,254,306,test/val_iceplains/2/000024.mp4,test/val_iceplains/2/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +255,255,10,test/val_desert/2/000003.mp4,test/val_desert/2/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +256,256,173,test/val_savanna/0/000020.mp4,test/val_savanna/0/000020,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +257,257,320,test/val_iceplains/1/000032.mp4,test/val_iceplains/1/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +258,258,115,test/val_desert/1/000044.mp4,test/val_desert/1/000044,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +259,259,341,test/val_iceplains/1/000038.mp4,test/val_iceplains/1/000038,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +260,260,61,test/val_desert/3/000024.mp4,test/val_desert/3/000024,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +261,261,231,test/val_savanna/1/000046.mp4,test/val_savanna/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +262,262,224,test/val_savanna/3/000042.mp4,test/val_savanna/3/000042,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +263,263,292,test/val_iceplains/3/000018.mp4,test/val_iceplains/3/000018,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +264,264,334,test/val_iceplains/1/000036.mp4,test/val_iceplains/1/000036,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +265,265,159,test/val_savanna/0/000013.mp4,test/val_savanna/0/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +266,266,102,test/val_desert/0/000039.mp4,test/val_desert/0/000039,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +267,267,362,test/val_iceplains/1/000046.mp4,test/val_iceplains/1/000046,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +268,268,322,test/val_iceplains/3/000032.mp4,test/val_iceplains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +269,269,381,test/val_plains/1/000004.mp4,test/val_plains/1/000004,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +270,270,84,test/val_desert/3/000032.mp4,test/val_desert/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +271,271,243,test/val_iceplains/1/000000.mp4,test/val_iceplains/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +272,272,65,test/val_desert/3/000025.mp4,test/val_desert/3/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +273,273,64,test/val_desert/0/000025.mp4,test/val_desert/0/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +274,274,272,test/val_iceplains/2/000010.mp4,test/val_iceplains/2/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +275,275,399,test/val_plains/0/000011.mp4,test/val_plains/0/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +276,276,248,test/val_iceplains/3/000001.mp4,test/val_iceplains/3/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +277,277,43,test/val_desert/3/000017.mp4,test/val_desert/3/000017,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +278,278,393,test/val_plains/2/000008.mp4,test/val_plains/2/000008,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +279,279,123,test/val_desert/2/000047.mp4,test/val_desert/2/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +280,280,29,test/val_desert/2/000011.mp4,test/val_desert/2/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +281,281,122,test/val_desert/0/000047.mp4,test/val_desert/0/000047,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +282,282,157,test/val_savanna/3/000011.mp4,test/val_savanna/3/000011,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +283,283,333,test/val_iceplains/2/000035.mp4,test/val_iceplains/2/000035,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +284,284,307,test/val_iceplains/1/000025.mp4,test/val_iceplains/1/000025,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +285,285,9,test/val_desert/0/000003.mp4,test/val_desert/0/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +286,286,250,test/val_iceplains/2/000002.mp4,test/val_iceplains/2/000002,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +287,287,165,test/val_savanna/3/000015.mp4,test/val_savanna/3/000015,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +288,288,0,test/val_desert/1/000000.mp4,test/val_desert/1/000000,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +289,289,271,test/val_iceplains/3/000010.mp4,test/val_iceplains/3/000010,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +290,290,276,test/val_iceplains/1/000013.mp4,test/val_iceplains/1/000013,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +291,291,82,test/val_desert/1/000031.mp4,test/val_desert/1/000031,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +292,292,191,test/val_savanna/2/000029.mp4,test/val_savanna/2/000029,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +293,293,374,test/val_plains/1/000001.mp4,test/val_plains/1/000001,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +294,294,446,test/val_plains/3/000032.mp4,test/val_plains/3/000032,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +295,295,378,test/val_plains/1/000003.mp4,test/val_plains/1/000003,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +296,296,152,test/val_savanna/0/000009.mp4,test/val_savanna/0/000009,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +297,297,54,test/val_desert/0/000021.mp4,test/val_desert/0/000021,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +298,298,259,test/val_iceplains/2/000005.mp4,test/val_iceplains/2/000005,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 +299,299,91,test/val_desert/2/000034.mp4,test/val_desert/2/000034,100,100,700,700,1200,1100,600,500,1,1501,1501,1501,1501 diff --git a/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.sha256 b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.sha256 new file mode 100644 index 0000000000000000000000000000000000000000..065350069e5b0f46e33cd51ad63f60aa956822fd --- /dev/null +++ b/outputs/dememwm_sourcefree_long_horizon/dememwm_sourcefree_ref_step200k_minecraft_test300_ctx600_roll500_bs1_seed43/test_manifest.sha256 @@ -0,0 +1 @@ +ef4e4de9d3975d8a7e724bb609e57f20186cd1f7285457c00fb85c1aca973b99